Mirror FTP Sites

Transcription

Mirror FTP Sites

R
Intel 810 Chipset Family
Programmer’s Reference Manual
November 1999
Revision 1.0
Order Number: 298026-001
®
Intel 810 Chipset Family PRM
THIS DOCUMENT IS PROVIDED "AS IS," WITH NO WARRANTIES WHATSOEVER, INCLUDING ANY
WARRANTY OF MERCHANTABILITY, NONINFRINGEMENT, FITNESS FOR ANY PARTICULAR PURPOSE,
OR ANY WARRANTY OTHERWISE ARISING OUT OF ANY PROPOSAL, SPECIFICATION OR SAMPLE. Intel®
disclaims all liability, including liability for infringement of any proprietary rights, relating to use of information in this
specification.
Information in this document is provided in connection with Intel products. No license, express or implied, by estoppel
or otherwise, to any intellectual property rights is granted by this document. Except as provided in Intel's Terms and
Conditions of Sale for such products, Intel assumes no liability whatsoever, and Intel disclaims any express or implied
warranty relating to sale and/or use of Intel products, including liability or warranties relating to fitness for a particular
purpose, merchantability or infringement of any patent, copyright or other intellectual property right. Intel products are
not intended for use in medical, lifesaving, or life-sustaining applications.
Intel may make changes to specifications and product descriptions at any time, without notice.
Designers must not rely on the absence or characteristics of any features or instructions marked “reserved” or
“undefined.” Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or
incompatibilities arising from future changes to them.
The Intel® 810 family of chipsets may contain design defects or errors known as errata that may cause the product to
deviate from published specifications. Current characterized errata are available upon request.
Copyright © Intel Corporation, 1999
* Third-party brands and names are the property of their respective owners.
ii
R
®
R
Contents
1. Introduction............................................................................................................................................... 1
1.1 Audience .............................................................................................................................................. 1
1.2 Reference Documents ......................................................................................................................... 1
®
1.3 Intel 82810 Chipset System ............................................................................................................... 2
®
®
2. Product Features of Intel 82810 Chipset & Intel 82810-DC100 Chipset .......................................... 5
2.1 GMCH Overview .................................................................................................................................. 9
3. System Address Map ............................................................................................................................. 11
3.1 GC Register Memory Address Map ................................................................................................... 15
3.2 Graphics Address Translation............................................................................................................ 18
3.3 Instruction Parser ............................................................................................................................... 19
3.4 Ring Buffers (RB) ............................................................................................................................... 20
3.4.1 Ring Buffer Registers.................................................................................................................... 21
3.4.2 Ring Buffer Initialization ................................................................................................................ 22
3.4.3 Ring Buffer Use............................................................................................................................. 22
3.5 Batch Buffers...................................................................................................................................... 23
4. Graphics Translation Table Range Definition ..................................................................................... 25
5. Basic Initialization Procedures ............................................................................................................. 27
5.1 Initialization Sequence ....................................................................................................................... 27
5.2 Hardware Detection (Probe) .............................................................................................................. 27
5.3 Frame Buffer Initialization .................................................................................................................. 29
5.4 Hardware Register Initialization ......................................................................................................... 29
5.4.1 Color vs. Monochrome Monitors ................................................................................................... 29
5.4.2 Protect Registers: Locking and Unlocking .................................................................................... 30
5.4.3 Checking Memory Frequency ....................................................................................................... 30
5.5 Hardware State .................................................................................................................................. 31
5.6 Saving the Hardware State ................................................................................................................ 31
5.7 Restoring the Hardware State............................................................................................................ 32
6. BLT Engine Programming ..................................................................................................................... 37
6.1 BLT Engine Programming Considerations......................................................................................... 37
6.1.1 When the Source and Destination Locations Overlap .................................................................. 37
6.2 Basic Graphics Data Considerations ................................................................................................. 41
6.2.1 Contiguous vs. Discontinuous Graphics Data .............................................................................. 41
6.2.2 Source Data .................................................................................................................................. 42
6.2.3 Monochrome Source Data ............................................................................................................ 43
6.2.4 Pattern Data .................................................................................................................................. 44
6.2.5 Destination Data............................................................................................................................ 47
6.3 BLT Programming Examples ............................................................................................................. 48
6.3.1 Pattern Fill — A Very Simple BLT................................................................................................. 48
6.3.2 Drawing Characters Using a Font Stored in System Memory ...................................................... 51
7. Initialization Registers ........................................................................................................................... 55
7.1 Standard VGA Registers.................................................................................................................... 55
7.2 SMRAM Registers.............................................................................................................................. 55
7.2.1 SMRAM—System Management RAM Control Register (Device 0) ............................................. 55
7.3 Graphics Controller Registers ............................................................................................................ 57
7.3.1 GR10Address Mapping ............................................................................................................. 57
7.3.2 GR11Page Selector .................................................................................................................. 60
7.4 CRT Controller Register..................................................................................................................... 61
7.4.1 CR30Extended Vertical Total Register ..................................................................................... 61
iii
®
7.4.2 CR31Extended Vertical Display End Register .......................................................................... 62
7.4.3 CR32Extended Vertical Sync Start Register............................................................................. 63
7.4.4 CR33Extended Vertical Blanking Start Register ....................................................................... 64
7.4.5 CR35Extended Horizontal Total Time Register ........................................................................ 65
7.4.6 CR39Extended Horizontal Blank Time Register ....................................................................... 65
7.4.7 CR40Extended Start Address Register..................................................................................... 66
7.4.8 CR41Extended Offset Register ................................................................................................. 67
7.4.9 CR42Extended Start Address High Register ............................................................................ 67
7.4.10 CR70Interlace Control Register .............................................................................................. 68
7.4.11 CR80I/O Control ...................................................................................................................... 68
7.4.12 CR82Blink Rate Control .......................................................................................................... 69
7.5 Display Control Registers................................................................................................................... 70
7.5.1 FW_BLC—FIFO Watermark and Burst Length Control................................................................ 70
7.6 I/O Control Registers.......................................................................................................................... 72
7.6.1 HVSYNC—HSYNC/VSYNC Control Register .............................................................................. 72
7.7 GPIO Registers .................................................................................................................................. 73
7.7.1 GPIOAGeneral-Purpose I/O Control Register A ....................................................................... 73
7.7.2 GPIOBGeneral-Purpose I/O Control Register B ....................................................................... 75
7.8 Clock Control Registers ..................................................................................................................... 77
7.8.1 Programming Notes ...................................................................................................................... 77
7.8.2 DCLK_0D—Display Clock 0 Divisor Register............................................................................... 79
7.8.5 LCD_CLKD—LCD Clock Divisor Register.................................................................................... 82
7.8.6 DCLK_0DS—Display & LCD Clock Divisor Select Register......................................................... 83
7.8.7 PWR_CLKC—Power Management and Miscellaneous Clock Control ........................................ 85
7.9 LCD / TV-Out Registers ..................................................................................................................... 86
7.9.1 HTOTAL—Horizontal Total Register............................................................................................. 86
7.9.2 HBLANK—Horizontal Blank Register ........................................................................................... 87
7.9.3 HSYNC—Horizontal Sync Register .............................................................................................. 88
7.9.4 VTOTAL—Vertical Total Register ................................................................................................. 88
7.9.5 VBLANK—Vertical Blank Register................................................................................................ 89
7.9.6 VSYNC—Vertical Sync Register................................................................................................... 90
7.9.7 LCDTV_C—LCD/TV-Out Control Register ................................................................................... 91
7.9.8 OVRACT—Overlay Active Register.............................................................................................. 94
7.9.9 BCLRPAT— Border Color Pattern Register ................................................................................. 94
7.10 Pixel Pipeline Control Registers ........................................................................................................ 95
7.10.1 PIXCONF—Pixel Pipeline Configuration .................................................................................... 95
7.11 Initialization Values for VGA Registers.............................................................................................. 98
8. Frame Buffer Access............................................................................................................................ 101
9. VGA and Extended VGA Registers ..................................................................................................... 103
9.1 General Control & Status Registers................................................................................................. 103
9.1.1 ST00Input Status 0.................................................................................................................. 104
9.1.2 ST01Input Status 1.................................................................................................................. 105
9.1.3 FCRFeature Control ................................................................................................................ 106
9.1.4 MSRMiscellaneous Output...................................................................................................... 107
9.2 Sequencer Registers........................................................................................................................ 108
9.2.1 SRXSequencer Index.............................................................................................................. 108
9.2.2 SR00Sequencer Reset............................................................................................................ 109
9.2.3 SR01Clocking Mode................................................................................................................ 109
9.2.4 SR02Plane/Map Mask............................................................................................................. 110
9.2.5 SR03Character Font ............................................................................................................... 111
iv
R
R
®
9.2.6 SR04Memory Mode Register .................................................................................................. 112
9.2.7 SR07Horizontal Character Counter Reset .............................................................................. 113
9.3 Graphics Controller Registers .......................................................................................................... 113
9.3.1 GRXGRX Graphics Controller Index Register ........................................................................ 113
9.3.2 GR00Set/Reset Register......................................................................................................... 114
9.3.3 GR01Enable Set/Reset Register............................................................................................. 114
9.3.4 GR02Color Compare Register ................................................................................................ 115
9.3.5 GR03Data Rotate Register ..................................................................................................... 115
9.3.6 GR04Read Plane Select Register........................................................................................... 116
9.3.7 GR05Graphics Mode Register ................................................................................................ 116
9.3.8 GR06Miscellaneous Register.................................................................................................. 119
9.3.9 GR07Color Don’t Care Register.............................................................................................. 120
9.3.10 GR08Bit Mask Register ......................................................................................................... 120
9.3.11 GR10Address Mapping ......................................................................................................... 121
9.3.12 GR11Page Selector............................................................................................................... 123
9.3.13 GR[14:1F]Software Flags x ................................................................................................... 123
9.4 Attribute Controller Registers ........................................................................................................... 124
9.4.1 ARXAttribute Controller Index Register................................................................................... 124
9.4.2 AR[00:0F]Palette Registers [0:F]............................................................................................. 125
9.4.3 AR10Mode Control Register ................................................................................................... 126
9.4.4 AR11Overscan Color Register ................................................................................................ 127
9.4.5 AR12Memory Plane Enable Register ..................................................................................... 128
9.4.6 AR13Horizontal Pixel Panning Register.................................................................................. 129
9.4.7 AR14Color Select Register ..................................................................................................... 130
9.5 VGA Color Palette Registers ........................................................................................................... 131
9.5.1 DACMASKPixel Data Mask Register ...................................................................................... 132
9.5.2 DACSTATEDAC State Register.............................................................................................. 132
9.5.3 DACRXPalette Read Index Register....................................................................................... 133
9.5.4 DACWXPalette Write Index Register ...................................................................................... 133
9.5.5 DACDATAPalette Data Register ............................................................................................. 133
9.6 CRT Controller Register................................................................................................................... 134
9.6.1 CRXCRT Controller Index Register ........................................................................................ 135
9.6.2 CR00Horizontal Total Register................................................................................................ 135
9.6.3 CR01Horizontal Display Enable End Register ........................................................................ 135
9.6.4 CR02Horizontal Blanking Start Register ................................................................................. 136
9.6.5 CR03Horizontal Blanking End Register .................................................................................. 136
9.6.6 CR04Horizontal Sync Start Register ....................................................................................... 137
9.6.7 CR05Horizontal Sync End Register ........................................................................................ 137
9.6.8 CR06Vertical Total Register.................................................................................................... 138
9.6.9 CR07Overflow Register........................................................................................................... 138
9.6.10 CR08Preset Row Scan Register ........................................................................................... 142
9.6.11 CR09Maximum Scan Line Register ...................................................................................... 143
9.6.12 CR0AText Cursor Start Register ........................................................................................... 144
9.6.13 CR0BText Cursor End Register ............................................................................................ 145
9.6.14 CR0CStart Address High Register ........................................................................................ 146
9.6.15 CR0DStart Address Low Register ......................................................................................... 147
9.6.16 CR0EText Cursor Location High Register............................................................................. 147
9.6.17 CR0FText Cursor Location Low Register.............................................................................. 148
9.6.18 CR10Vertical Sync Start Register ......................................................................................... 148
9.6.19 CR11Vertical Sync End Register .......................................................................................... 149
9.6.20 CR12Vertical Display Enable End Register .......................................................................... 150
v
®
9.6.21
9.6.22
9.6.23
9.6.24
9.6.25
9.6.26
9.6.27
9.6.28
9.6.29
9.6.30
9.6.31
9.6.32
9.6.33
9.6.34
9.6.35
9.6.36
9.6.37
9.6.38
9.6.39
9.6.40
9.6.41
9.6.42
CR13Offset Register ............................................................................................................. 150
CR14Underline Location Register ......................................................................................... 151
CR15Vertical Blanking Start Register ................................................................................... 152
CR16Vertical Blanking End Register..................................................................................... 152
CR17CRT Mode Control ....................................................................................................... 153
CR18Line Compare Register ................................................................................................ 156
CR22Memory Read Latch Data Register.............................................................................. 156
CR24 Test Register for Toggle State of Attribute Controller Register .................................. 157
CR30Extended Vertical Total Register.................................................................................. 157
CR31Extended Vertical Display End Register ...................................................................... 158
CR32Extended Vertical Sync Start Register ......................................................................... 159
CR33Extended Vertical Blanking Start Register ................................................................... 160
CR34 Extended Vertical Blank Time Register (Reserved, Not Implemented).......................... 161
CR35 Extended Horizontal Total Time Register ................................................................... 161
CR39Extended Horizontal Blank Time Register ................................................................... 161
CR40Extended Start Address Register................................................................................. 161
CR41Extended Offset Register ............................................................................................. 163
CR42Extended Start Address High Register ........................................................................ 163
CR70Interlace Control Register ............................................................................................ 164
CR80I/O Control .................................................................................................................... 164
CR81Reserved ...................................................................................................................... 165
CR82Blink Rate Control ........................................................................................................ 165
10. Programming Interface ..................................................................................................................... 167
10.1 Reserved Bits and Software Compatibility ...................................................................................... 167
10.2 Overview.......................................................................................................................................... 167
10.3 GC Register Programming .............................................................................................................. 168
10.4 GC Instruction Streams ................................................................................................................... 169
10.4.1 Instruction Use .......................................................................................................................... 169
10.4.2 Instruction Transport Overview ................................................................................................. 169
10.4.3 Instruction Parser ...................................................................................................................... 169
10.4.4 Ring Buffers (RB) ...................................................................................................................... 170
10.4.5 Batch Buffers............................................................................................................................. 173
10.4.6 Instruction Arbitration ................................................................................................................ 173
10.5 Instruction Format............................................................................................................................ 176
10.5.1 Instruction Parser Instructions................................................................................................... 176
10.5.2 2D Instructions .......................................................................................................................... 176
10.5.3 3D Instructions .......................................................................................................................... 176
11. Instruction Parser Instructions ........................................................................................................ 181
11.1 Introduction ...................................................................................................................................... 181
11.2 Instruction Descriptions ................................................................................................................... 181
11.2.1 GFXCMDPARSER_NOP_IDENTIFICATION ........................................................................... 181
11.2.2 GFXCMDPARSER_BREAKPOINT_INTERRUPT.................................................................... 182
11.2.3 GFXCMDPARSER_USER_INTERRUPT ................................................................................. 182
11.2.4 GFXCMDPARSER_WAIT_FOR_EVENT ................................................................................. 183
11.2.5 GFXCMDPARSER_FLUSH ...................................................................................................... 184
11.2.6 GFXCMDPARSER_CONTEXT_SEL........................................................................................ 185
11.2.7 GFXCMDPARSER_DEST_BUFFER_INFO ............................................................................. 186
11.2.8 GFXCMDPARSER_FRONT_BUFFER_INFO .......................................................................... 187
11.2.9 GFXCMDPARSER_Z_BUFFER_INFO .................................................................................... 188
11.2.10 GFXCMDPARSER_REPORT_HEAD..................................................................................... 189
11.2.11 GFXCMDPARSER_ARB_ON_OFF........................................................................................ 189
11.2.12 GFXCMDPARSER_OVERLAY_FLIP ..................................................................................... 190
vi
R
R
11.2.13
11.2.14
11.2.15
11.2.16
11.2.17
®
GFXCMDPARSER_LOAD_SCAN_LINES_INCL ................................................................... 190
GFXCMDPARSER_LOAD_SCAN_LINES_EXCL .................................................................. 191
GFXCMDPARSER_STORE_DWORD_IMM .......................................................................... 191
GFXCMDPARSER_STORE_DWORD_INDEX ...................................................................... 192
GFXCMDPARSER_BATCH_BUFFER ................................................................................... 193
12. 2D Instructions .................................................................................................................................. 195
12.1 Introduction ...................................................................................................................................... 195
12.2 BLTs to and from Cacheable Memory............................................................................................. 195
12.3 BLT Engine Instructions .................................................................................................................. 196
12.3.1 SETUP_BLT.............................................................................................................................. 196
12.3.2 SETUP_MONO_PATTERN_SL_BLT ....................................................................................... 198
12.3.3 PIXEL_BLT ............................................................................................................................... 199
12.3.4 SCANLINE_BLT........................................................................................................................ 200
12.3.5 TEXT_BLT ................................................................................................................................ 201
12.3.6 TEXT_Immediate_BLT.............................................................................................................. 202
12.3.7 COLOR_BLT ............................................................................................................................. 203
12.3.8 PAT_BLT................................................................................................................................... 204
12.3.9 MONO_PAT_BLT ..................................................................................................................... 205
12.3.10 SRC_COPY_BLT .................................................................................................................... 206
12.3.11 SRC_COPY_IMMEDIATE_BLT .............................................................................................. 207
12.3.12 MONO_SRC_COPY_BLT....................................................................................................... 208
12.3.13 MONO_SRC_COPY_IMMEDIATE_BLT................................................................................. 210
12.3.14 FULL_BLT ............................................................................................................................... 212
12.3.15 FULL_MONO_SRC_BLT ........................................................................................................ 213
12.3.16 FULL_MONO_PATTERN_BLT............................................................................................... 215
12.3.17 FULL_MONO_PATTERN_MONO_SRC_BLT ........................................................................ 217
12.4 BLT Engine Instruction Definitions .................................................................................................. 219
12.4.1 BR00—BLT Opcode & Control ................................................................................................. 219
12.4.2 BR01—Setup BLT Raster OP, Control, and Destination Offset ............................................... 222
12.4.3 BR02—Clip Rectangle Y1 Address .......................................................................................... 224
12.4.4 BR03—Clip Rectangle Y2 Address .......................................................................................... 224
12.4.5 BR04—Clip Rectangle X1 and X2 ............................................................................................ 225
12.4.6 BR05—Setup Expansion Background Color ............................................................................ 225
12.4.7 BR06—Setup Expansion Foreground Color ............................................................................. 226
12.4.8 BR07—Setup Color Pattern Address........................................................................................ 226
12.4.9 BR08—Destination X1 and X2.................................................................................................. 227
12.4.10 BR09—Destination Address and Destination Y1 Address...................................................... 228
12.4.11 BR10—Destination Y2 Address .............................................................................................. 228
12.4.12 BR11—BLT Source Pitch (Offset) or Monochrome Source Quadwords ................................ 229
12.4.13 BR12—Source Address .......................................................................................................... 230
12.4.14 BR13—BLT Raster OP, Control, and Destination Pitch ......................................................... 230
12.4.15 BR14—Destination Width & Height......................................................................................... 232
12.4.16 BR15—Color Pattern Address ................................................................................................ 233
12.4.17 BR16—Pattern Expansion Background & Solid Pattern Color ............................................... 234
12.4.18 BR17—Pattern Expansion Foreground Color ......................................................................... 234
12.4.19 BR18—Source Expansion Background, and Destination Color ............................................. 235
12.4.20 BR19—Source Expansion Foreground Color ......................................................................... 235
12.4.21 S_SLADD—Source Scan Line Address.................................................................................. 236
12.4.22 D_SLH—Destination Scan Line Height................................................................................... 236
12.4.23 D_SLRADD—Destination Scan Line Read Address .............................................................. 237
13. Rendering Engine Instructions ........................................................................................................ 239
13.1 GFXPRIMITIVE ............................................................................................................................... 239
13.1.1 Axis-Aligned Rectangles ........................................................................................................... 239
vii
®
13.1.2 Primitive Winding Order ............................................................................................................ 239
13.1.3 Position Mask ............................................................................................................................ 240
13.1.4 Bias ........................................................................................................................................... 240
13.2 Primitive Rendering Instruction Format ........................................................................................... 240
13.2.1 Variable-Length Vertex Formats for Rendering Instructions..................................................... 241
13.3 GFXVERTEX ................................................................................................................................... 242
13.4 GFXRENDERSTATE_VERTEX_FORMAT..................................................................................... 243
13.4.1 Non-Pipelined State Variables .................................................................................................. 244
13.5 GFXRENDERSTATE_MAP_TEXELS............................................................................................. 244
13.6 GFXRENDERSTATE_MAP_COORD_SETS.................................................................................. 246
13.7 GFXRENDERSTATE_MAP_INFO .................................................................................................. 248
13.8 GFXRENDERSTATE_MAP_FILTER .............................................................................................. 254
13.9 GFXRENDERSTATE_MAP_LOD_LIMITS ..................................................................................... 256
13.10 GFXRENDERSTATE_MAP_LOD_CONTROL.............................................................................. 257
13.11 GFXRENDERSTATE_MAP_PALETTE_LOAD............................................................................. 258
13.12 GFXRENDERSTATE_MAP_COLOR_BLEND_STAGES ............................................................. 259
13.13 GFXRENDERSTATE_MAP_ALPHA_BLEND_STAGES .............................................................. 262
13.14 GFXRENDERSTATE_COLOR_FACTOR..................................................................................... 264
13.15 GFXRENDERSTATE_COLOR_CHROMA_KEY .......................................................................... 265
13.16 GFXRENDERSTATE_SRC_DST_BLEND_MONO ...................................................................... 266
13.17 GFXRENDERSTATE_Z_BIAS_ALPHA_FUNC_REF................................................................... 270
13.18 GFXRENDERSTATE_LINE_WIDTH_CULL_SHADE_MODE ...................................................... 271
13.19 GFXRENDERSTATE_BOOLEAN_ENA_1.................................................................................... 273
13.20 GFXRENDERSTATE_BOOLEAN_ENA_2.................................................................................... 274
13.21 GFXRENDERSTATE_FOG_COLOR............................................................................................ 275
13.22 GFXRENDERSTATE_DRAWING_RECTANGLE_INFO .............................................................. 276
13.23 GFXRENDERSTATE_SCISSOR_ENABLE .................................................................................. 277
13.24 GFXRENDERSTATE_SCISSOR_RECTANGLE_INFO ............................................................... 278
13.25 Stipple Pattern ............................................................................................................................... 279
13.26 GFXRENDERSTATE_ANTI_ALIASING........................................................................................ 279
13.27 GFXRENDERSTATE_PROVOKING_VTX_PIXELIZATION_RULE ............................................. 281
13.28 GFXRENDERSTATE_DEST_BUFFER_VARIABLES .................................................................. 283
14. Clock Control Registers ................................................................................................................... 285
14.1 Programming Notes......................................................................................................................... 285
14.2 DCLK_0D — Display Clock 0 Divisor Register ............................................................................... 287
14.5 LCD_CLKD — LCD Clock Divisor Register .................................................................................... 290
14.6 DCLK_0DS — Display & LCD Clock Divisor Select Register ......................................................... 291
14.7 PWR_CLKC — Power Management and Miscellaneous Clock Control ......................................... 294
15. Video Registers ................................................................................................................................. 297
15.1 OV0ADD—Overlay 0 Register Update Address Register............................................................... 299
15.2 DOV0STA—Display/Overlay 0 Status Register .............................................................................. 300
15.3 Gamma Correction .......................................................................................................................... 302
15.3.1 GAMC[5:0]—Gamma Correction Registers .............................................................................. 302
15.3.2 Mathematical Gamma Correction for Overlay........................................................................... 304
15.4 Overlay Buffer Pointer Registers ..................................................................................................... 306
15.4.1 OBUF_0Y—Overlay Buffer 0 Y Pointer Register...................................................................... 307
15.4.2 OBUF_1Y—Overlay Buffer 1 Y Pointer Register...................................................................... 307
15.4.3 OBUF_0U—Overlay Buffer 0 U Pointer Register ..................................................................... 307
15.4.4 OBUF_0V—Overlay Buffer 0 V Pointer Register...................................................................... 308
15.4.5 OBUF_1U—Overlay Buffer 1 U Pointer Register ..................................................................... 309
15.4.6 OBUF_1V—Overlay Buffer 1 V Pointer Register...................................................................... 309
viii
R
R
®
15.5 Overlay Stride Registers.................................................................................................................. 310
15.5.1 OV0STRIDE—Overlay 0 Stride Register.................................................................................. 310
15.6 Overlay Initial Phase Registers ....................................................................................................... 311
15.6.1 YRGB_VPH—Y/RGB Vertical-Phase Register......................................................................... 311
15.6.2 UV_VPH—UV Vertical-Phase Register .................................................................................... 312
15.6.3 HORZ_PH—Horizontal-Phase Register ................................................................................... 312
15.6.4 INIT_PH—Initial Phase Register............................................................................................... 313
15.7 Overlay Destination Window Position/Size Registers ..................................................................... 314
15.7.1 DWINPOS—Destination Window Position Register ................................................................. 314
15.7.2 DWINSZ—Destination Window Size Register .......................................................................... 315
15.8 Overlay Source Size Registers........................................................................................................ 316
15.8.1 SWID—Source Width Register ................................................................................................. 316
15.8.2 SWIDQW—Source Width in Qwords Register.......................................................................... 317
15.8.3 SHEIGHT—Source Height Register ......................................................................................... 318
15.9 Overlay Scale Factor Registers....................................................................................................... 319
15.9.1 YRGBSCALE—Y/RGB Scale Factor Register.......................................................................... 319
15.9.2 UVSCALE—U V Scale Factor Register .................................................................................... 320
15.10 Overlay Color Correction Registers............................................................................................... 321
15.10.1 OV0CLRC0—Overlay 0 Color Correction 0 Register ............................................................. 321
15.10.2 OV0CLRC1—Overlay 0 Color Correction 1 Register ............................................................. 321
15.11 Overlay Destination Color Key Registers ...................................................................................... 322
15.11.1 DCLRKV—Destination Color Key Value Register................................................................... 322
15.11.2 DCLRKM—Destination Color Key Mask Register................................................................... 322
15.12 Overlay Source Color Key Registers............................................................................................. 323
15.12.1 SCLRKVH—Source Color Key Value High Register .............................................................. 324
15.12.2 SCLRKVL—Source Color Key Value Low Register................................................................ 324
15.12.3 SCLRKM—Source Color Key Mask Register ......................................................................... 325
15.13 Overlay Configuration Registers.................................................................................................... 326
15.13.1 OV0CONF—Overlay Configuration Register .......................................................................... 326
15.14 OV0CMD—Overlay Command Register ....................................................................................... 327
15.15 Overlay Alpha Blend Window Position/Size Registers.................................................................. 331
15.15.1 AWINPOS—Alpha Blend Window Position Register .............................................................. 331
15.15.2 AWINSZ—Alpha Blend Window Size Register ....................................................................... 332
15.16 Overlay Flip Instruction .................................................................................................................. 332
16. Instruction and Interrupt Control Registers ................................................................................... 333
16.1 Instruction Control Registers ........................................................................................................... 333
16.1.1 FENCE—Graphics Memory Fence Table Registers ................................................................ 333
16.1.2 PGTBL_CTL—Page Table Control Register ............................................................................ 335
16.1.3 PGTBL_ER—Page Table Error Register (Debug).................................................................... 337
16.1.4 RINGBUF—Ring Buffer Registers ............................................................................................ 338
16.1.5 HWS_PGA—Hardware Status Page Address Register ........................................................... 339
16.1.6 IPEIR—Instruction Parser Error Identification Register (Debug) .............................................. 340
16.1.7 IPEHR—Instruction Parser Error Header Register (Debug)..................................................... 340
16.1.8 INSTDONE—Instruction Stream Interface Done Register ....................................................... 341
16.1.9 NOPID—NOP Identification Register........................................................................................ 342
16.1.10 INSTPM—Instruction Parser Mode Register .......................................................................... 342
16.1.11 INSTPS—Instruction Parser State Register (Debug) ............................................................. 344
16.1.12 BBP_PTR—Batch Buffer Parser Pointer Register (Debug).................................................... 344
16.1.13 ABB_STR—Active Batch Buffer Start Address Register (Debug) .......................................... 345
16.1.14 ABB_END—Active Batch Buffer End Address Register (Debug) ........................................... 346
16.1.15 DMA_FADD—DMA Engine Fetch Address (Debug) .............................................................. 347
16.2 Interrupt Control Registers .............................................................................................................. 348
16.2.1 HWSTAM—Hardware Status Mask Register............................................................................ 350
16.2.2 IER—Interrupt Enable Register................................................................................................. 351
ix
®
16.2.3
16.2.4
16.2.5
16.2.6
IIR—Interrupt Identity Register.................................................................................................. 352
IMR—Interrupt Mask Register................................................................................................... 353
ISR—Interrupt Status Register.................................................................................................. 354
Error Identity, Mask and Status Registers ................................................................................ 355
17. LCD / TV-Out Register Description.................................................................................................. 359
17.1 HTOTAL — Horizontal Total Register ............................................................................................. 359
17.2 HBLANK — Horizontal Blank Register............................................................................................ 360
17.3 HSYNC — Horizontal Sync Register............................................................................................... 360
17.4 VTOTAL — Vertical Total Register ................................................................................................. 361
17.5 VBLANK — Vertical Blank Register ................................................................................................ 362
17.6 VSYNC — Vertical Sync Register ................................................................................................... 363
17.7 LCDTV_C — LCD/TV-Out Control Register.................................................................................... 364
17.8 OVRACT — Overlay Active Register .............................................................................................. 367
17.9 BCLRPAT — Border Color Pattern Register................................................................................... 368
17.10 Reserved Registers ....................................................................................................................... 368
18. Local Memory Interface .................................................................................................................... 369
18.1 DRT—DRAM Row Type.................................................................................................................. 369
18.2 DRAMCL—DRAM Control Low ....................................................................................................... 370
18.3 DRAMCH—DRAM Control High...................................................................................................... 371
19. I/O Control Registers ........................................................................................................................ 373
19.1 HVSYNC—HSYNC/VSYNC Control Register................................................................................. 373
19.2 GPIO Registers ............................................................................................................................... 374
19.2.1 GPIOAGeneral-Purpose I/O Control Register A ................................................................... 374
19.2.2 GPIOBGeneral-Purpose I/O Control Register B ................................................................... 376
20. Display and Cursor Registers .......................................................................................................... 379
20.1 DISP_SL—Display Scan Line Count............................................................................................... 379
20.2 DISP_SLC—Display Scan Line Count Range Compare ................................................................ 380
20.3 Pixel Pipeline Control ...................................................................................................................... 381
20.3.1 PIXCONF—Pixel Pipeline Configuration .................................................................................. 381
20.4 BLTCNTL—BLT Control.................................................................................................................. 384
20.5 Reserved Register ........................................................................................................................... 385
20.6 SWF[1:3]—Software Flag Registers ............................................................................................... 385
20.6.1 DPLYBASE—Display Base Address Register.......................................................................... 386
20.6.2 DPLYSTAS—Display Status Select Register ........................................................................... 387
20.7 Hardware Cursor ............................................................................................................................. 389
20.7.1 CURCNTR—Cursor Control Register ....................................................................................... 389
20.7.2 CURBASE—Cursor Base Address Register ............................................................................ 390
20.7.3 CURPOS—Cursor Position Register ........................................................................................ 391
21.
x
Mode Parameters .............................................................................................................................. 393
R
R
®
Figures
Figure 1. GMCH simplified block diagram ........................................................................................................ 3
®
®
Figure 2. Intel 82810 chipset system block diagram with Intel 82810 GMCH and either ICH or ICH0 ......... 7
®
®
Figure 3. Intel 82810 chipset system block diagram with Intel 82810-DC100 GMCH and ICH .................... 8
Figure 4. GMCH block diagram ........................................................................................................................ 9
Figure 5. System memory map ....................................................................................................................... 12
Figure 6. Graphics controller I/O and memory map........................................................................................ 13
Figure 7. GTT mapping ................................................................................................................................... 19
Figure 8. Graphics controller instruction interface .......................................................................................... 20
Figure 9. Ring buffers...................................................................................................................................... 20
Figure 10. Batch buffer sequence ................................................................................................................... 23
Figure 11. Memory with overlay active ........................................................................................................... 24
Figure 12. Source corruption in BLT with overlapping source and destination locations ............................... 38
Figure 13. Correctly performed BLT with overlapping source and destination locations................................ 39
Figure 14. Suggested starting points for possible source and destination overlaps....................................... 40
Figure 15. Representation of on-screen single 6-pixel line in the frame buffer .............................................. 41
Figure 16. Representation of on-screen 6×4 array of pixels in the frame buffer ............................................ 42
Figure 17. Pattern data (always an 8×8 array of pixels) ................................................................................. 44
Figure 18. 8-bpp pattern data — Occupies 64 bytes (8 quadwords) .............................................................. 45
Figure 19. 16-bpp pattern data — Occupies 128 bytes (16 quadwords) ........................................................ 45
Figure 22. On-screen destination for example pattern fill BLT ....................................................................... 48
Figure 23. Pattern data for example pattern fill BLT ....................................................................................... 49
Figure 24. Results of example pattern fill BLT ................................................................................................ 50
Figure 25. On-screen destination for example character drawing BLT .......................................................... 51
Figure 26. Source data in system memory for example character drawing BLT............................................ 51
Figure 27. Results of example character-drawing BLT................................................................................... 53
Figure 28. Graphics controller I/O and memory map.................................................................................... 168
Figure 29. Graphics controller instruction interface ...................................................................................... 170
Figure 30. Ring buffers.................................................................................................................................. 170
Figure 31. Batch buffer sequence ................................................................................................................. 173
Figure 32. Instruction format for first dword .................................................................................................. 176
Figure 33. Rectangle vertices ....................................................................................................................... 239
Figure 34. State variable relationships.......................................................................................................... 244
xi
®
Tables
Table 1. Memory-Mapped Registers............................................................................................................... 15
Table 2. Ring Buffer Characteristics ............................................................................................................... 21
Table 3. VGA Address Range......................................................................................................................... 59
Table 4. CRT Display Sync Polarities ........................................................................................................... 108
Table 5. VGA Address Range....................................................................................................................... 122
Table 6. Memory Address Counter Address Bits [15:0]................................................................................ 155
Table 7. Frame Buffer Address Decoder ...................................................................................................... 155
Table 8. Ring Buffer Characteristics ............................................................................................................. 172
Table 9. Graphics Controller Instructions...................................................................................................... 177
Table 10. Summary of Source Surface Formats with Filter Output Channel Mappings ............................... 248
Table 11. Overlay Register/Instruction Categories ....................................................................................... 298
Table 12. Bit Definition for Interrupt Control Registers ................................................................................. 348
xii
R
R
®
1. Introduction
The Intel® 82810 chipset is a highly integrated chipset designed for the basic graphics/multimedia PC platform. The
chipset consists of a Graphics and Memory Controller Hub (GMCH) host bridge and an I/O Controller Hub (ICH/ICH0)
bridge for the I/O subsystem. The GMCH integrates a system memory DRAM controller that supports a 64-bit, 100MHz DRAM array. The DRAM controller is optimized for maximum efficiency.
There are two versions of the GMCH (i.e., 82810, 82810-DC100), which are pin compatible. The difference between the
two versions is that the Intel® 82810-DC100 integrates a display cache DRAM controller that supports a 4-MB, 32-bit,
100-MHz DRAM array for enhanced 2D and 3D performance.
This document describes both versions of the GMCH (i.e., 82810, 82810-DC100).
An overview of the Intel 82810 chipset is provided in the next section.
Notes: In this document “GMCH” refers to both the 82810 and 82810-DC100 chipsets, unless otherwise specified. The
Intel 82810 and Intel 82810-DC100 chipsets may contain design defects or errors known as errata, which may
cause the product to deviate from published specifications. Current characterized errata are available upon
request.
1.1
Audience
This document is intended for hardware, software, and firmware designers who seek to implement or utilize the graphic
functions of the Intel 82810 and Intel 82810-DC100 chipsets. Familiarity with 2D and 3D graphics programming is
assumed.
1.2
Reference Documents
The following documents should be available for reference when using this specification:
• Inte1® 82810/82810-DC100 Graphics and Memory Controller Hub (GMCH) Datasheet
• Intel® 82801AA (ICH) and Intel® 82801AB (ICH0) I/O Controller Hub Datasheet
• Intel® 82802AB/82802AC Firmware Hub (FWH) Datasheet
1
®
1.3
Intel® 82810 Chipset System
The Intel 82810 Chipset uses a hub architecture, with the GMCH as the host bridge hub and the 82801xx I/O Controller
Hub (ICH) as the I/O hub. The ICH is a highly integrated, multifunctional I/O controller hub that provides the interface
to the PCI bus and integrates many of the functions needed in today’s PC platforms. The GMCH and ICH communicate
over a dedicated hub interface. As for the GMCH, there are two versions of the ICH (i.e., 82801AA: ICH, 82801AB:
ICH0). This provides added flexibility in designing cost-effective system solutions. These devices are pin compatible
and are housed in 241-pin packages. The GMCH devices are designed to work with either ICH or ICH0.
82801AA (ICH) / 82801AB (ICH0) functions and capabilities include:
•
PCI Rev. 2.2 compliance, with support for 33-MHz PCI operations
•
ICH0 supports up to 4 Req/Gnt pairs (PCI slots). ICH supports up to 6 Req/Gnt pairs (PCI slots).
•
Power management logic support
•
Enhanced DMA controller, interrupt controller, and timer functions
•
Integrated IDE controller. ICH0 supports Ultra ATA/33. ICH also supports Ultra ATA/66.
•
USB host interface, with support for 2 USB ports
•
System Management Bus (SMBus) compatible with most I2C devices
•
AC ’97 2.1-compliant link for audio and telephony CODECs
•
Low Pin Count (LPC) interface
•
Firmware Hub (FWH) interface support
•
Alert On LAN* (82801AA ICH only)
2
R
®
R
Figure 1. GMCH simplified block diagram
HA[31:3]#
HD[63:0]#
ADS#
BNR#
BPRI#
DBSY#
DEFER#
DRDY#
HIT#
HITM#
HLOCK#
HREQ[4:0]#
HTRDY#
RS[2:0]#
CPURST#
SMAA[11:0]
SMAB[7:4]#
SBS[1:0]
SMD[63:0]
SDQM[7:0]
SCS[3:0]#
SRAS#
SCAS#
SWE#
SCKE[1:0]
LCS#
LDQM[3:0]#
LSRAS#
LSCAS#
LMA[11:0]
LWE#
LMD[31:0]
Display
Interface
System Bus
Interface
Digital
TV
Out
System
Memory
Interface
Display
Cache
Interface
(Intel® 82810DC100
Chipset only)
VSYNC
HSYNC
IREF
RED
GREEN
BLUE
DDCSCL
DDCSDA
LTVCL
LTVDA
TVCLKIN/INT#
CLKOUT[1:0]
BLANK#
LTVDATA[11:0]
TVSYNC
TVHSYNC
Clock
Signals
HCLK
SCLK
LTCLK
LOCLK
LRCLK
DCLKREF
HLCLK
Misc.
Interface
Signals
GLRREFA
GTLREFB
RESET#
Hub
Interface
HUBREF
HL[10:0]
HLSTRB
HLSTRB#
HCOMP
Note: The Display Cache Interface is only avaliable on the
Intel® 82810-DC100 chipset.
gmch_blk.vsd
3
®
R
This page intentionally left blank.
4
R
®
2. Product Features of Intel® 82810 Chipset &
Intel® 82810-DC100 Chipset
5
®
R
Processor/Host Bus Support
2D Graphics
•
•
•
•
•
•
•
•
• Hardware-accelerated functions
• 3-operand raster BitBLTs
• 64×64×3-color transparent cursor
Optimized for the Intel® CeleronTM processor
Supports processor 370-pin socket connector
Supports 32-bit system bus addressing
4-deep in-order queue; 4- or 1-deep request queue
Supports uniprocessor systems only
In-order & dynamic deferred transaction support
66/100-MHz system bus frequency
AGTL+ I/O buffer
Arithmetic Stretch Blitter Video
• NTSC and PAL TV-out support
• H/W overlay engine with bilinear filtering
• Independent gamma correction, saturation,
brightness, and contrast for overlay
Integrated DRAM Controller
• 8 MB to 256 MB using 16-Mb/64-Mb technology
(512 MB using 128-Mb technology)
• Supports up to 2 double-sided DIMM modules
• 64-bit data interface
• 100-MHz system memory bus frequency
• Support for asymmetrical DRAM addressing only
• Support for ×8, ×16 and ×32 DRAM device width
• Refresh mechanism: CBR ONLY supported
• Enhanced open page arbitration SDRAM paging
scheme
• Suspend to RAM support
Integrated Graphics Controller
•
•
•
•
3D hyper-pipelined architecture
Parallel data processing (PDP)
Precise pixel interpolation (PPI)
2D H/W acceleration
3D Graphics Visual Enhancements
•
•
•
•
•
•
Flat & Gouraud shading
Bilinear and anisotropic filtering
Fogging atmospheric effects
Z-buffering
3D pipe 2D clipping
Backface culling
3D Graphics Texturing Enhancements
• Per-pixel perspective correction texture mapping
• Texture compositing
®
Display Cache I/F (Intel 82810-DC100 chipset only)
• 32-bit data interface
• 100-MHz SDRAM interface
• Support for 1M×16, (4 MB only)
Arbitration Scheme and Concurrency
• Centralized arbitration model for optimum
concurrency support
• Concurrent operations of processor and system
busses, supported via dedicated arbitration and data
buffering
Data Buffering
• Distributed data buffering model for optimum
concurrency
• DRAM write buffer with read-around-write
capability
• Dedicated CPU-DRAM, hub interface-DRAM, and
graphics-DRAM read buffers
Power Management Functions
• SMRAM space remapping to A0000h (128 KB)
• Optional extended SMRAM space above 256 MB;
additional 512-KB/1-MB TSEG from top of memory;
cacheable
• Stop clock grant and halt special cycle translation
from the host to the hub interface
• ACPI-compliant power management
• APIC buffer management
• SMI, SCI, and SERR error indication
Display
• Integrated 24-bit, 230-MHz RAMDAC
• Gamma-corrected video
• DDC2B compliant
Supporting I/O Bridge
• 241-pin BGA I/O controller hub (ICH0/ICH)
Packaging/Power
• 421 BGA
• 1.8V core with 3.3V CMOS I/O
6
®
R
®
®
Figure 2. Intel 82810 chipset system block diagram with Intel 82810 GMCH and either ICH or ICH0
®
Intel® Pentium III Processor,
Intel® Pentium® II Processor,
and
Intel® Celeron™ Processor
System Bus (66/100 MHz)
Digital Video Out
Intel® 810 Chipset
Encoder
Intel® 82810 Chipset
(GMCH0)
TV
Display
64 Bit /
100 MHz Only
- Memory Controller
- Graphics Controller
- 3D Engine
- 2D Engine
- Video Engine
System
Memory
PCI Slots
(ICH = 6 Req/Gnt pairs)
(ICH0 = 4 Req/Gnt pairs)
PCI Bus
2 IDE Ports
ICH = Ultra ATA/66
ICH0 = Ultra ATA/33
ICH and ICH0
(I/O Controller Hub)
ISA
Option
Super
I/O
LAN
Option
AC'97
2 USB
Ports
USB
Audio Codec
USB
Modem Codec
FWH
(Firmware Hub)
sysblk1.vsd
7
®
R
®
®
Figure 3. Intel 82810 chipset system block diagram with Intel 82810-DC100 GMCH and ICH
Intel® Pentium® lll processor,
Intel® Pentium® II processor,
and
Intel ® Celeron™ Processor
System Bus (66/100 MHz)
Digital Video Out
Intel® 810 Chipset
Encoder
Intel® 82810-DC100
Chipset (GMCH)
TV
Display
Display Cache
64 Bit /
100 MHz Only
- Memory Controller
- Graphics Controller
- 3D Engine
- 2D Engine
- Video Engine
(4-MB SDRAM,
100 MHz Only)
System
Memory
PCI Slots
(ICH=6 Req/Gnt pairs)
PCI Bus
ICH
(I/O Controller Hub)
2 IDE Ports
Ultra ATA/66
ISA
Option
Super
I/O
LAN
Option
AC'97
2 USB
Ports
USB
Audio Codec
USB
Modem Codec
FWH
(Firmware Hub)
sysblk2.vsd
8
®
R
2.1
GMCH Overview
Figure 4 is a block diagram of the GMCH, illustrating the various interfaces and integrated components of the GMCH
chip.
GMCH functions and capabilities include:
•
Support for a single-processor configuration
•
64-bit GTL+-based system bus interface at 66 MHz / 100 MHz
•
32-bit host address support
•
64-bit system memory interface with optimized support for SDRAM at 100 MHz
•
Integrated 2D & 3D graphics engines
•
Integrated 230-MHz DAC
•
Integrated digital video out port
•
4-MB display cache (82810-DC100 only)
Figure 4. GMCH block diagram
System Bus Interface
Buffer
Analog
Display
Out
Display Engine
3D Engine
HW Motion Comp
3D
Engine
DAC
Overlay
DDC
I2C
Memory Interface
2D Engine
HW Cursor
Digital
Video
Out
System
Memory
Digital Video Out
Port
Display Cache
Memory
(Intel® 82810DC100
Chipset only)
Stretch
BLT Eng
BLT Eng
Buffer
Hub Interface
gmch_blk2.vsd
9
®
R
10
R
®
3. System Address Map
The following figure shows the system memory address map for the Intel® 82810 chipset. The GMCH memory map
includes a number of programmable ranges. ALL of these ranges must be unique and non-overlapping. There are NO
hardware interlocks to prevent problems in the case of overlapping ranges. Accesses to overlapped ranges may produce
indeterminate results.
11
®
R
System Memory Space
64 GB
Extended CPU
Memory Space
4 GB
PCI Memory
FEEC 0000h
FEEB FFFFh
Optional HSEG
FEEA 0000h
FEE9 FFFFh
PCI Memory
Size=0/512 KB (fixed)
Graphics Controller (GC)
PCI Memory accesses
(memory-mapped
to GC registers
control/status registers)
Base=MMADR Reg.
(14h); Dev 1
PCI Memory
Size=0/32/64MB; MISCC Reg.
(72h); Dev 0
Logical Graphics
Memory
(32MB/64MB)
PCI Memory Accesses to
Graphics Logical Memory
Translated Through
Graphics Translation
Table (GTT)
Base=GMADR Reg.
(10h); Dev 1
PCI Memory
TOM (512 MB Max.)
Size=0KB/512KB/1MB;
SMRAM Reg. (70h); Dev 0
0FFFFFh
Optional TSEG
Segment F
(BIOS Shadow Area, etc.)
Size=0KB/512KB/1MB;
Optional Graphics Buffer
SMRAM Reg. (70h); Dev 0
Main
Memory
Size=64KB GFX Trans. Tbl (GTT)GTT Access via GC
Control Registers
Base Programmed in GC Unit
960 KB
Segment E
0E0000h
0DFFFFh
896 KB
Optional CD Hole
880 KB
Segment D
0D0000h (BIOS Shadow Area, etc.)
832 KB
0CFFFFh
Optional ISA Hole
15 MB
C0000h
BFFFFh
0F0000h
0EFFFFh
0DC000h
0DBFFFh
16 MB
100000h
0FFFFFh
1 MB
Segment C
1 MB
0C0000h
0BFFFFh
Video BIOS
(shadowed in memory)
Graphics Adapter
(128 KB)
A0000h
9FFFFh
DOS
Compatibility
Memory
Optionally
mapped to the
internal GC
0A0000h
09FFFFh
640 KB
DOS Area
(640 KB)
System/Application SW
00000h
768 KB
Std PCI/ISA Video Mem
(SMM Mem) 128 KB
000000h
0 KB
mem_map.vsd
Figure 5. System memory map
12
®
R
Except for the PCI Configuration registers, all of the GC registers are memory mapped. The base address of this 512-KB
memory block is programmed in the MMADR PCI configuration register. The following figure shows the high-level
memory map of the GC registers. Note that 2D control registers (VGA and extended VGA registers) also are located at
their standard I/O locations.
Memory Space Map
(512 KB allocation)
- Cursor Registers
- Display Registers
- Pixel Pipe Registers
Offset From
Base_Reg
7FFFFh
70000h
6FFFFh
Reserved
60000h
5FFFFh
- Video Capture Registers
- DVD Registers
50000h
4FFFFh
Blt Engine Control Status (RO)
40000h
3FFFFh
Overlay Registers1
30000h
2FFFFh
Note:
1. Some Overlay registers are doublebuffered with an additional address range
in graphics memory. See Overlay Register
Chapter for details.
Reserved
20000h
1FFFFh
Page Table Range
10000h
0FFFFh
Reserved
07000h
06FFFh
Clock Control Registers
06000h
05FFFh
Misc I/O Control Registers
05000h
04FFFh
Reserved
04000h
03FFFh
Local Memory Interface
Control Registers
I/O Space Map
(Standard graphics locations)
VGA and Ext. VGA Registers
- Instruction Control Regs.
- Fence Table Registers
- Interrupt Control
03000h
02FFFh
01000h
00FFFh
31
00000h
19
MMADR Register
(Base Address)
reginstm.vsd
.
Figure 6. Graphics controller I/O and memory map
13
®
R
VGA and Extended VGA Control Registers (00000h−
−00FFFh)
These registers are located in both I/O space and memory space. The VGA and extended VGA registers contain the
following register sets: General Control/Status, Sequencer (SRxx), Graphics Controller (GRxx), Attribute Controller
(ARxx), VGA Color Palette, and CRT Controller (CRxx) registers. Detailed bit descriptions are provided in the
chapter that discusses the VGA and extended VGA registers. The registers within a set are accessed using an indirect
addressing mechanism, as described at the beginning of each section. Note that some register description sections
have additional operational information at the beginning of the section.
Instruction, Memory, and Interrupt Control Registers (01000h−
−02FFFh)
The instruction and interrupt control registers are located in main memory space and contain the following types of
registers:
• Instruction Control Registers
Ring buffer registers and page table control registers are located in this address range. Various instruction status,
error, and operating registers are located in this group of registers.
• Graphics Memory Fence Registers
The graphics memory fence registers are used for memory tiling capabilities.
• Interrupt Control/Status Registers
This register set provides interrupt control/status for various GC functions.
• Display Interface Control Register
This register controls the FIFO watermark and provides burst length control.
Local Memory Registers (03000h−
−03FFFh)
These registers are located in main memory space and provide local memory DRAM control.
I/O Control Registers (05000h−
−05FFFh)
This chapter provides I/O control register functions.
Clock Control Registers (06000h−
−06FFFh)
This memory address space is the location of the GC clock control and power management registers.
Page Table Range (10000h−
−1FFFFh)
Overlay Registers (30000h−
−3FFFFh)
These registers provide control of the GC overlay engine. The overlay registers are double-buffered, with one register
buffer located in graphics memory and the other on the GC chip. On-chip registers are not directly writeable. To
update the on-chip registers, software writes to the register buffer area in graphics memory and instructs the GC to
update the on-chip registers.
Blitter Status Registers (40000h−
−4FFFFh)
For debug purposes only, a set of read-only registers provides visibility into the BLT engine status.
LCD/TV-out, and HW DVD Registers (60000h−
−6FFFFh)
This memory address range is used for LCD/TV-out control registers and for HW DVD control registers.
Cursor, Display, and Pixel Pipe Registers (70000h−
−7FFFFh)
This memory address range is used for cursor control, display, and pixel pipe control registers.
14
®
R
3.1
GC Register Memory Address Map
All GC registers are memory-mapped. In addition, the VGA and extended VGA registers are I/O mapped.
Table 1. Memory-Mapped Registers
Address Offset
Symbol
00000h−00FFFh

01000h−01FFFh
02000h−0201Fh

FENCE[0:7]
02020h−02023h
02024h−02027h
02030h–0207Fh
02030h–0203Fh
02040h–0204Fh
02050h–0207Fh
02080h–02083h
02084h–02087h
02088h–0208Bh
0208Ch–0208Fh
02090h–02091h
02092h–02093h
02094h–02097h
02098h−002099h
0209Ah–0209Fh
020A0h−020A1h
020A2h–020A3h
020A4h−020A5h
020A6h–020A7h
020A8h−020A9h
020AAh–020ABh
020ACh−020ADh
Register Name
VGA and VGA Extended Registers
These registers are both memory- and I/O-mapped and are listed in the
following table. Note that the I/O address and memory offset address
are the same value for each register.
Instruction and Interrupt Control Registers (01000h−02FFFh)
Reserved
Access

Graphics Memory Fence Table Register [0:7]

R/W
PGTBL_CTL
Page Table Control Register
R/W
PGTBL_ER
Page Table Error Register
RO
RINGBUF
Ring Buffer Registers
Low Priority Ring Buffer (4 DWs)
Interrupt Ring Buffer (4 DWs)
Reserved
Hardware Status Page Address Register
Reserved
R/W
HWS_PGA

IPEIR
IPEHR
INSTDONE

NOPID
HWSTAM

IER
Instruction Parser Error Identification Register
Instruction Parser Error Header Register
Instruction Stream Interface Done Register
Reserved
NOP Identification Register
Hardware Status Mask Register
Reserved
Interrupt Enable Register

IIR
Interrupt Identity Register

IMR
Interrupt Mask Register

ISR
Interrupt Status Register
Reserved
Reserved
Reserved
R/W

RO
RO
RO

RO
R/W

R/W

R/WC

R/W

RO

EIR
Reserved

EMR
Reserved
Error Mask Register
020B8h−020B9h

ESR
Error Status Register

RO
020BAh−020BBh

Reserved

020BDh−020BFh
020C0h
020C1h–020C3h

INSTPM
Reserved

R/W
020AEh−020AFh
020B0h−020B1h
020B2h−020B3h
020B4h−020B5h
020B6h−020B7h
020C4h–020C7h
020C8h–020CBh
020CCh–020CFh

INSTPS
BBP_PTR
ABB_SRT
Error Identity Register
Reserved
Instruction Parser Mode Register
Reserved
Instruction Parser State Register
Batch Buffer Parser Pointer Register
Active Batch Buffer Start Address Register

R/WC

R/W

RO
RO
RO
15
®
R
Address Offset
020D0h–020D3h
020D4h–020D7h
020D8h–020DBh
020DCh–020DBh
Symbol
ABB_END
DMA_FADD
FW_BLC
Register Name
Active Batch Buffer End Address Register
DMA Engine Fetch Address Register
FIFO Watermark and Burst Length Control
Reserved
020DCh–020DFh

MEM_MODE
020E0h−02FFFh

03000h
03001h
03002h
DRT
DRAMCL
DRAMCH
03003h−03FFFh

Memory Control Registers (03000h−03FFFh)
DRAM Row Type
DRAM Control Low
DRAM Control High
Reserved
04000h−04003h

Span Cursor Registers (04000h−04FFFh)
Reserved
04004h−04007h
04008h−0400Bh

UI_SC_CTL
0400Ch−04FFFh

Memory Interface Mode Register
Reserved
Reserved
Span Cursor Control
Reserved
I/O Control Registers (05000h−05FFFh)
HSYNC/VSYNC Control
Access
RO
RO
R/W

R/W

R/W
R/W
R/W



R/W

05000h−05003h
HVSYNC
05010h−05013h
GPIOA
General-Purpose I/O Control A
R/W
05014h−05017h
GPIOB
General-Purpose I/O Control B
R/W
05018h−05FFFh

Reserved
06000h−06003h
Clock Control and Power Management Registers (06000h−06FFFh)
DCLK_0D
Display Clock 0 Divisor
R/W
06004h−06007h
DCLK_1D
R/W
06008h−0600Bh
DCLK_2D
R/W
0600Ch−0600Fh
LCD_CLKD
LCD Clock Divisor
R/W
06010h−06013h
DCLK_0DS
Display and LCD Clock Divisor Select
R/W
06014h−06017h
PWR_CLKC
Power Management and Miscellaneous Clock Control
R/W
07000h−0FFFFh

Reserved (07000h−0FFFFh)
Reserved
10000h−1FFFFh
Graphics Translation Table Range Definition (10000h−1FFFFh)
GTT
Graphics Translation Table Range Definition
20000h−2FFFFh

Reserved (20000h−2FFFFh)
Reserved
Overlay Registers (30000h−03FFFFh)
Overlay 0 Register Update Address Overlay 0
R/W


WO

30000h−30003h
OV0ADDR
30004h−30007h
—
Reserved
—
30008h−3000Bh
DOV0STA
Display/Overlay 0 Status
RO
3000Ch−3000Fh
—
Reserved
30010h−30027h
GAMMA[5:0]
Gamma Correction [5:0] (6 registers)
30028h−300FFh
30100h–30103
—
Reserved
—
OBUF_0Y
OBUF_1Y
Overlay Buffer 0 Y Pointer
RO
RO
30104h−30107h
R/W
—
R/W
30108h−3010Bh
OBUF_0U
Overlay Buffer 0 U Pointer
RO
3010Ch−3010Fh
OBUF_0V
Overlay Buffer 0 V Pointer
RO
30110h−30113h
OBUF_1U
RO
30114h−30117h
OBUF_1V
RO
16
®
R
30118h−3011Bh
Address Offset
Symbol
OV0STRIDE
Overlay 0 Stride
Register Name
Access
RO
3011Ch−3011Fh
YRGB_VPH
Y/RGB Vertical Phase
RO
30120h−30123h
UV_VPH
UV Vertical Phase
RO
30124h−30127h
HORZ_PH
Horizontal Phase
RO
30128h−3012Bh
INIT_PH
Initial Phase
RO
3012Ch−3012Fh
DWINPOS
Destination Window Position
RO
30130h−30133h
DWINSZ
Destination Window Size
RO
30134h−30137h
SWID
Source Width
RO
30138h−3013Bh
SWIDQW
Source Width In qwords
RO
3013Ch−3013Fh
SHEIGHT
Source Height
RO
30140h−30143h
YRGBSCALE
Y/RGB Scale Factor
RO
30144h−30147h
UVSCALE
U V Scale Factor
RO
30148h−3014Bh
OV0CLRC0
Overlay 0 Color Correction 0
RO
3014Ch−3014Fh
OV0CLRC1
Overlay 0 Color Correction 1
RO
30150h−30153h
DCLRKV
Destination Color Key Value
RO
30154h−30157h
DCLRKM
Destination Color Key Mask
RO
30158h−3015Bh
SCLRKVH
Source Color Key Value High
RO
3015Ch−3015Fh
SCLRKVL
Source Color Key Value Low
RO
30160h−30163h
SCLRKM
Source Color Key Mask
RO
30164h−30167h
OV0CONF
Overlay 0 Configuration
RO
30168h−3016Bh
OV0CMD
Overlay 0 Command
RO
30170h−30173h
AWINPOS
Alpha Blend Window Position
RO
30174h−30177h
AWINZ
Alpha Blend Window Size
RO
30178h−3FFFFh

Reserved

40000h–40003h
40004h–40007h
40008h–4000Bh
4000Ch–4000Fh
40010h–40013h
40014h–40017h
40018h–4001Bh
4001Ch–4001Fh
40020h–40023h
40024h–40027h
40028h–4002Bh
4002Ch–4002Fh
40030h–40033h
40034h–40037h
40038h–4003Bh
4003Ch–4003Fh
40040h–40043h
40044h–40047h
40048h–4004Bh
4004Ch–4004Fh
40074h–40077h
40078h–4007Bh
4007Ch–4007Fh
BLT Engine Status (40000h−4FFFFh) (Software Debug)
BR00
BLT Opcode and Control
BR01
Setup BLT Raster OP, Control, and Destination Offset
BR02
Clip Rectangle Y1 Address
BR03
Clip Rectangle Y1 Address
BR04
Clip Rectangle X1 and X2 Address
BR05
Setup Expansion Background Color
BR06
Setup Expansion Foreground Color
BR07
Setup Color Pattern Address
BR08
Destination X1 and X2
BR09
Destination Address and Destination Y1 Address
BR10
Destination Y2 Address
BR11
BLT Source Pitch (Offset) or Monochrome Source Quadwords
BR12
Source Address
BR13
BLT Raster OP Control, and Destination Pitch
BR14
Destination Width and Height
BR15
Color Pattern Address
BR16
Pattern Expansion Background and Solid Pattern Color
BR17
Pattern Expansion Foreground Color
BR18
Source Expansion Background and Destination Color
BR19
Source Expansion Foreground Color
SSLADD
Source Scan Line Address
DSLH
Destination Scan Line Height
DSLRADD
Destination Scan Line Read Address
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
RO
17
®
R
Address Offset
40080h–4FFFFh
Symbol

Register Name
Reserved
Access

LCD/TV-Out and HW DVD Registers (60000h–6FFFFh)
LCD/TV-Out
60000h–60003h
60004h–60007h
60008h–6000Bh
6000Ch–6000Fh
60010h–60013h
60014h–60017h
60018h–6001Bh
6001Ch–6001Fh
60020h–60023h
HTOTAL
HBLANK
HSYNC
VTOTAL
VBLANK
VSYNC
LCDTV_C
OVRACT
BCLRPAT
70000h–70003h
70004h–70007h
70008h–7000Bh
7000Ch–7000Fh
70014h–7001Fh
70020h–70023h
70024h–70027h
70080h–70083h
70084h–70087h
70028h–7002Bh
7002Ch–7FFFFh
Display and Cursor Control Registers (70000h–7FFFFh)
DISP_SL
Display Scan Line Count
DISP_SLC
Display Scan Line Count Range Compare
PIXCONF
Pixel Pipeline Configuration
BLTCNTL
BLT Control
SWF[1:3]
Software Flags [1:3] (3 registers)
DPLYBASE
Display Base Address
DPLYSTAS
Display Status Select
CURCNTR
Cursor Control and Vertical Extension
CURBASE
Cursor Base Address
CURPOS
Cursor Position
Reserved

3.2
Horizontal Total
Horizontal Blank
Horizontal Sync
Vertical Total
Vertical Blank
Vertical Sync
LCD / TV-out Control
Overlay Active Register
Border Color Pattern
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W

Graphics Address Translation
The Intel 82180 chipset employs a logical memory addressing concept for accessing graphics data. The GC supports a
64-MB logical address space, where each 4-KB logical page can be mapped to a physical memory page in system RAM,
PCI memory or an optional display cache memory. This mapping is performed by means of a Graphics Translation Table
(GTT).
GC engines can address the full 64-MB logical address space. The CPU is provided access to either the full 64-MB space
or just the lower 32 MB, via a PCI memory range associated with the graphics device.
The GTT is allocated in system RAM and maintained by the graphics driver. The 4 KB-aligned physical address of the
64-KB GTT is programmed via the GC’s PGTBL register.
Each 16K-dword GTT entry can map a 4-KB logical page to a physical memory page. Fields in the GTT entry control
the mapping of that logical page in the following manner:
(GTT register field “V”) whether or not that logical 4-KB page is mapped to a physical memory page. Accesses to
invalid pages will result in an error interrupt.
(GTT register field “T1T0”) the physical memory address space of the mapped page:
System RAM page (no processor cache snoop)
PCI memory page (processor cache snooped if below TOM)
Display cache page
The page number of the mapped page (within the particular physical memory address space)
18
®
R
Although the GTT format permits any logical page to be mapped to any page in the supported physical memory address
spaces, the GC imposes restrictions on the how specific graphics operands (buffers, etc.) can be mapped to physical
memory.
The GTT entries must be written via a GTT alias in the graphics device’s memory-mapped register space (10000h1FFFFh). This allows the GC to snoop GTT entry writes and invalidate graphics TLBs as required. The GTT entries
must not be written directly in system memory.
System Memory
4 GB
31
0
Graphics
Translation Table
(GTT)
31
0
Base + 64 MB
Virtual
Graphics
Memory
Graphics Engine
Address Space
64 KB
31
0
64 MB
Base + 32 MB
Base
0
TOM
0 KB
Optional Display
Cache
4 KB
GTT Maps 4KB blocks of Virtual
Graphics Memory to 4 KB pages in
System Memory
4 KB
GTT Maps 4KB blocks of
Virtual Graphics Memory
to 4 KB pages in Display
Cache
gtt.vsd
Figure 7. GTT mapping
3.3
Instruction Parser
The GC's Instruction Parser (IP) unit is responsible for
•
Detecting the presence of instructions (within the ring buffers)
•
Arbitrating the execution of instruction streams
•
Reading instructions from ring buffers and batch buffers via DMA
•
Parsing the common "client" (destination) field of instructions
•
Executing instruction parser instructions (which control IP functionality, provide synchronization functions, and
provide miscellaneous GC control functions)
•
Redirecting 2D and 3D instructions to the appropriate destination, while following drawing engine concurrency and
coherency rules
Figure 8 is a high-level diagram of the GC instruction interface.
19
®
R
Low Priority Ring
(Graphics Memory)
Batch Buffers
Instruction
Batch Buff Instr
DMA
Instruction
Instruction access and decoding
2D Instructions
Instr
Parser
DMA
FIFO
3D Instructions (3D state,
3D Primitives, STRBLT,
Motion Compensation)
Interrupt Ring
(Graphics Memory)
Batch Buffers
BLT
Engine
Instruction
DMA
3D
Engine
Batch Buff Instr
Instruction
Instruction Parser Instructions
- Parser Control
(e.g., Batch Buffer Instr., NOP,
Sync ID, Flush, breakpoint )
-
Memory Interface Control
(e.g., Store DWord to memory)
- Display/Overlay Control
(e.g., Front Buffer, Scan
Lines, Overlay Flip
Display
Engine
Overlay
Engine
cmd_str.vsd
Figure 8. Graphics controller instruction interface
3.4
Ring Buffers (RB)
The GC provides two Ring Buffer (RB) mechanisms via which instructions can be passed to the instruction parser. They
are referred to as the interrupt and low-priority RBs, and they are basically identical, except for differences in arbitration
rules and priority.
Graphics Memor y
Wrap Around
Starting Address
Buff er
Length
Low Priorit y Ring
Buff er
Head Pointer
DMA Point er
Tail Pointer
Starting Address
Head Pointer
Buff er
Length
Interrupt Ring
Buff er
DMA Point er
Tail Pointer
RingBuf. vsd
Figure 9. Ring buffers
20
®
R
3.4.1
A ring buffer is defined by a set of four ring buffer registers. Before an RB can be used for instruction transport, software
needs to program these registers. The fields contained within these registers are as follows:
Ring Buffer Valid: This bit controls whether the RB is included in the instruction arbitration process. Software must
program all other RB parameters before enabling an RB. An RB can be disabled and later re-enabled. Enabling or
disabling an RB does not, of itself, change any other RB register fields.
Start Address: This field points to a contiguous, 4KB-aligned, linear (e.g., not tiled) memory address region, which
provides the actual instruction buffer area.
Buffer Length: The size of the buffer, in 4-KB increments, up to 2 MB.
Head Offset: This is the dword offset (from the start address) of the next instruction executed by the IP. The IP will
update this field as instructions are retired. (Note that, if instructions are pending execution, the IP likely will have
fetched instructions past the head offset). Because the GC does not “reset” the head offset when an RB is enabled,
software must program the head offset field before enabling the ring buffer. Although this allows software to enable
an RB with any legal values for head/tail (i.e., it can enable or re-enable the RB with instructions already pending),
the software is expected to initialize the head offset at 0. Once the head offset reaches the tail offset (head = tail), the
IP considers the RB "empty."
Head Wrap Count: This field is incremented by the IP each time the head offset wraps back to the start of the buffer.
As it is included in the dword written in the “report head” process, software can use this field to track IP progress as
if the RB had a “virtual” length 2048 times the size of the actual physical buffer.
Tail Offset: This is the qword offset (from the start address) where software will write the next instruction. After
writing instructions into the RB, software updates the tail offset field in order to submit the instructions for execution
(by setting it to the qword offset immediately following the last instruction to be submitted). The submitted
instructions can wrap from the end of the buffer back to the top, in which case the tail offset written will be less than
the previous value. Note that, because the RB empty condition is defined as "head offset == tail offset," software
must leave at least one qword free at all times. (That is, the buffer is considered "full" when only one qword is free.)
Automatic Report Head Enable: Software can request that the hardware head pointer register contents be written
("reported") periodically to snooped system memory. This is desirable, as software needs to use the head offset to
determine the amount of free space in the RB, and having the head pointer periodically reported to system memory
provides a fairly accurate head offset value automatically (i.e., without having to explicitly store a head offset value
via an instruction). The head pointer register will be stored at an RB-specific displacement into the "hardware status
page" (defined by the HWSTAM register).
Table 2. Ring Buffer Characteristics
Characteristic
Description
Alignment
4-KB-page aligned
Max. size
2 MB
Length
Programmable in number of 4-KB pages
Start pointer
Programmable page-aligned address of the buffer
Head pointer
Programmable to initially setup ring
Hardware-maintained dword offset in the ring buffer. Pointer wraps.
DMA pointer
Hardware-maintained DMA request double qword offset. Pointer wraps.
Tail pointer
Programmable double qword offset in the ring buffer
21
®
3.4.2
Ring Buffer Initialization
Before initializing an RB, software must first allocate the desired number of 4-KB pages for use as buffer space. Then
the RINGBUF registers associated with the RB are programmed. Once the Ring Buffer Valid bit is set, the RB will be
considered for instruction arbitration, and the head and tail offsets will either indicate an empty RB (i.e., head == tail) or
will define some number of instructions to be executed.
3.4.3
Ring Buffer Use
Software can write new instructions into the "free space" of the RB, from the tail offset up to (but not including) the
qword prior to the qword indicated by the head offset. (Remember, software must leave at least one qword empty in the
RB at all times.) Note that this "free space" may wrap from the end of the RB back to the start.
Software must use some mechanism to track instruction execution progress, in order to determine the "free space" in the
RB. This can either be:
A direct read of the Head Pointer register
The automatic reporting of the Head Pointer register
The explicit reporting of the Head Pointer register via the GFXCMDPARSER_REPORT_HEAD instruction
Another "implicit" means by which software can determine how far the IP has progressed in retiring instructions from
an RB. This could include the use of "Store DWORD" instructions to write sequencing data to system memory.
Once the instructions have been written (and padded out to a qword, if necessary), software can write the Tail Pointer
register to submit the new instructions for execution.
22
R
®
R
3.5
Batch Buffers
The GC provides for the execution of instruction sequences external to RBs. These sequences are called "batch buffers,"
and they are initiated through the use of GFXCMDPARSER_BATCH_BUFFER instructions that specify the starting
address and length of the batch buffers. The arbitration rules used by the IP when executing batch buffers differ from
those employed when executing RBs, and they are described later in this chapter. When a batch buffer instruction is
executed out of an RB, an initiated batch buffer sequence allows the GC to read the instructions sequentially (via DMA)
from the batch buffer.
What happens when the end of the batch buffer is reached depends on the final instruction in the buffer. If the final
instruction is a GFXCMDPARSER_BATCH_BUFFER instruction, another batch buffer sequence is initiated. This
process, called "chaining," continues until a batch buffer terminates with an instruction other than
GFXCMDPARSER_BATCH_BUFFER, at which point execution will resume in the RB at the instruction following the
initial GFXCMDPARSER_BATCH_BUFFER.
Instruction
From Ring Buff er
Instruction
Buff er
Chaining
Batch Buff Instr
Instruction
Instruction
Batch Buff Instr
Buff er
Chaining
Instruction
Instruction
Instruction
Return to Ring Buff er
btch_buf.vsd
Figure 10. Batch buffer sequence
23
®
R
System Memory
Graphics Controller
Register Range
(512 KB)
Base+301xxh
Overlay 0 Reg
On-chip registers for Overlay 0
(read only; debug)
(Base = MMADR PCI Reg.)
Graphics Memory
Base+xxh
(Base = OV0ADD Reg.)
Overlay 0 Reg
overlay1.vsd
Figure 11. Memory with overlay active
24
Memory buffer area for loading on-chip registers
- Software setsup register values
- HW updates on-chip regs for next VBLANK
®
R
4. Graphics Translation Table Range Definition
Address offset:
Default value:
Access:
10000h-1FFFFh
Page table range 64 KB
Aligned dword-qword, Write Only
This range defined within the graphics memory-mapped register space enables the memory manager to access the
graphics translation table. A page table write will invalidate that entry in internal translation table caches (TLBs). The
translation table resides in system memory and can be accessed directly by the memory manager. However, to ensure
coherency between hardware-maintained translation caches and the translation table in main memory, the memory
manager must use this range to update the translation table.
The page table must be QW aligned, with each entry being dword aligned such that each QW stores the translation for
two 4-KB pages. The page table base address for graphics memory will be programmed in the PGTB_CNTL register.
For a graphics memory of 64 MB with a TLB block size of 4 KB, 16 K entries will be needed. Each entry can be
accommodated in 4Bs, so the page table will be 64 KB in size.
Page Table Entry: 1 dword per 4-KB page
31
30
XX=00
V:
29
12
11
Physical address 29:12
3
Reserved
2
1
T1T0
0
V
1 = Valid page table entry (PTE)
0 = Invalid page table entry (PTE). An access to an invalid PTE will result in an interrupt.
T1T0:
01 = Physical address targets local memory.
00 = Physical address targets main memory (not snooped).
11 = Physical address targets cacheable main memory (results in snoop on processor bus) .
10 = Reserved.
Note:
T1T0 = 11 is used only if the surface is a blit source or destination operand used within the context of a
source copy command.
XX:
Reserved. Must be 0.
25
®
R
26
®
R
5. Basic Initialization Procedures
5.1
Initialization Sequence
The initialization of graphics driver resources can be broken down into three categories: hardware detection, frame
buffer initialization, and hardware register initialization. Each category is discussed in more detail in the following
sections.
In all discussions that follow, there is a basic assumption that the graphics adapter has completed the power-on video
BIOS initialization or video BIOS reset. Therefore, the adapter is in a known state and will respond in compliance with
the VGA and VESA specifications.
5.2
Hardware Detection (Probe)
Most operating systems will probe for installed devices. The Intel® 8281x family of devices advertises their presence in
PCI space by using unique values in the PCI VendorId and DeviceId locations. The following table lists the device IDs
used to identify the members of the 8281x family of graphics adapters:
Vendor Id
Device Id
PCI Device
PCI Offset: 0
PCI Offset: 2
Number
8086
7120
0
82810 bridge
8086
7121
1
82810
8086
7122
0
82810 DC100 bridge
8086
7123
1
82810 DC100
8086
7124
0
82810E bridge
8086
7125
1
82810E
8086
1100
0
82815 GMCH host-hub interface bridge /
DRAM controller FSB limited to 100 MHz
8086
1101
1
82815 FSB limited to 100 MHz
AGP bridge
8086
1102
2
82815 FSB limited to 100 MHz
Internal graphics device
8086
1110
0
82815 no AGP, internal graphics only
GMCH host-hub interface bridge / DRAM controller
8086
1112
2
82815 no AGP, internal graphics only
8086
1120
0
82815 no internal graphics, AGP only
8086
1121
1
No internal graphics, AGP only
AGP bridge
Characteristics
27
®
R
8086
1130
0
82815 fully featured Solano
8086
1131
1
AGP bridge
8086
1132
2
Once the operating system has identified the device, it can load the appropriate driver.
One of the first tasks of the driver is to make sure that the device matches the driver. Checking that the driver and device
match is done in much the same way that the operating system identifies the graphics adapter. That is, the PCI VendorId
and ProductId values are examined. Some operating systems will make available to the driver the values it found during
its scan. If not, the driver must scan the PCI space until it finds a match on the VendorId and ProductId values. The
driver normally caches this information so that it is accessible by other driver modules, when needed.
The next task of the device driver is to ensure that required resources are present. These resources include the minimum
memory requirements, IO address space requirements, and operating system support requirements (such as GART
support). If the driver detects that the operating system or the physical hardware does not meet the driver’s minimum
requirements, the driver should not load. The operating system should then be able to make use of the graphics adapter in
its VGA- and VESA-compliant mode.
If the operating system and hardware support are present, the driver should acquire the blocks of memory and IO address
space that will be required. These blocks should include at least the following:
28
•
Memory-mapped IO address space: 512 KB beginning at 0x80000
•
Linear frame buffer space: 32 or 64 MB beginning at 0xfe000000
•
Legacy IO addresses to support monochrome or color monitors
•
VGA IO addresses
®
R
5.3
Frame Buffer Initialization
The frame buffer initialization is responsible for setting up the memory that will contain the display data. Other objects
also can be stored in display memory.
The following steps should be performed:
5.4
•
Map a 0x80000-byte region in memory to the MMIO base address. The base address of the memorymapped region should be programmed into the MMADDR register, offset 14 in the PCI address space.
•
Allocate enough memory for the frame buffer from a memory pool created during initialization. The
amount of memory is determined by system characteristics, but should default to at least 8 MB.
•
If a hardware cursor is being used, allocate memory for the hardware cursor from the same memory
pool. The hardware does not use the GART to access the memory for the cursor, so local-to-physical
memory address translation must be performed. The hardware cursor memory address should be
programmed into the CURBASE register, memory-mapped address 70084h.
•
The low-priority ring buffer memory should be initialized to 0. The low-priority ring buffer pointers
should be programmed into the ring buffer pointer registers, RINGBUF, which begin at offset 2030h in
the memory-mapped IO space.
Hardware Register Initialization
5.4.1
Color vs. Monochrome Monitors
The mapping and initialization of some hardware registers depends in part on whether the graphics adapter is attached to
a monochrome or color monitor. The following steps illustrate how to determine the type of output device attached to the
graphics adapter:
•
Read the Miscellaneous Output Register (0x3cc).
•
Test the low-order bit of the Miscellaneous Output Register, and interpret it as follows:
•
0: The adapter is in monochrome monitor mode. In this mode, the control register is 3b4 and
3b5, and status is at 3ba.
•
1: The adapter is in color monitor mode. In this mode, the control register is 3d4 and 3d5, and
status is at 3da.
See the section on VGA compatibility for a description of the register space that must be acquired.
29
®
5.4.2
Protect Registers: Locking and Unlocking
In order to make use of some protected VGA registers, a locking and unlocking mechanism needs to be implemented.
The following steps illustrate how to unlock (or unprotect) the VGA registers:
•
Send a VERT_SYNC_END value to the register at vgaBase + 4.
•
Read the value in the register at vgaBase + 5.
•
Clear the high-order bit of the value just read.
•
Write the resulting value back into the register at vgaBase + 5.
5.4.3
Checking Memory Frequency
The driver behavior occasionally must be modified, depending on the frequency at which the memory is running. The
following steps illustrate how to determine the local memory frequency:
30
•
Read the contents of the Intel 82810 Chipset Configuration Register (PCI address space 0x50).
•
Examine the value of bit 4.
•
The value is interpreted as follows:
•
0: Frequency is 100 MHz.
•
1: Frequency is 133 MHz.
R
®
R
5.5
Hardware State
Under certain conditions, it may be necessary to save and restore the hardware state of the graphics adapter. These
conditions include mode switching, output device switching, processing changes in power state, and others. The next two
sections provide a brief description of the state saving and restoration requirements.
5.6
Saving the Hardware State
Note that the VGA register unlocking protocol must be performed in order to access some of the registers described
below.
The driver should preserve the following registers during a state change in order to provide complete state restoration in
the future:
IO Control
CR80
Address Mapping
GR10
Bit Blit Control
MM 0x7000c
Video Clock 2 / M
MM 0x6008
Video Clock 2 / N
MM 0x600c
Video Clock 2 / Divisor Select
MM 0x6012
Vertical Total
CRX 30
Vertical Display End
CRX 31
Vertical Sync Start
CRX 32
Vertical Blank Start
CRX 33
Horizontal Total
CRX 35
Horizontal Blnk
CRX 39
Ext Offset
CRX 41
Interlace Control
CRX 70
MM 0x2098
MM 0x20a0
Interrupt Identity Register
MM 0x20a4
MM 0x20a8
Error Mask Register
MM 0x20b4
Display Control Register
MM 0x70008
Pixel Pipeline Configuration 0 Register
MM 0x70009
31
®
5.7
R
MM 0x7000a
MM 0x7000b
Watermark and Burstlength Control
MM 0x20d8
Low-priority ring information
MM 0x2030 – 0x203f
Restoring the Hardware State
The graphics adapter state should be restored by performing the following steps. Note some of the synchronization
operations, especially those that ensure that the local memory is idle during the state restore. Also, much of the work
involves reprogramming the registers with the values captured during the save-state operation.
•
Blank the screen.
•
Turn off DRAM refresh.
Read the value of the DRAM_CONTROL_HI Register (MM 0x3002).
•
Set the DRAM Refresh Rate bits (DDR Bits 4:3) to Disable_Refresh (value 0).
•
Write the modified value back to the DRAM_CONTROL_HI Register.
•
Write the M, N, and P (i.e., the Divisor Select value) values from the saved state information.
•
Restore the 8-bit DAC mode to what it was when the state was saved, but preserve the current value of
the rest of the register containing this flag:
•
32
•
•
Read the Pixel Pipeline Configuration 0 Register.
•
Clear the current value of the 8- or 6-bit DAC mode.
•
OR–in (only) the value of the DAC_8_BIT from saved register information of the Pixel
Pipeline Configuration 0 Register.
•
Write the result back to the Pixel Pipeline Configuration 0 Register.
Restore the generic VGA registers to the values captured at save-state time.
®
R
•
•
Restore the following registers to their saved state values:
Vertical Total
CRX 30
Vertical Display End
CRX 31
Vertical Sync Start
CRX 32
CRX 33
Horizontal Total
CRX 35
Horizontal Blnk
CRX 39
Ext Offset
CRX 41
The following registers should restore only certain bits from the saved state values:
Interlace Control
•
Read the current value.
•
Clear the interlace enable bit.
•
OR–in the saved value of the Interlace Control Register.
•
Write the result back into the Interlace Control Register.
Address Mapping:
•
CRX 70
GR10
•
Read the current value of the Address Mapping Register.
•
Save only the reserved bits values (bits 7:5).
•
OR–in the saved value of the Address Mapping Register.
•
Write the result back into the Address Mapping Register.
Now the DRAM refresh can be turned on:
•
Read the value of the DRAM_CONTROL_HI Register.
•
Turn off the DRAM_REFRESH_RATE bits.
•
OR–in a 60-Hz refresh rate value.
•
Write the result back into the DRAM_CONTROL_HI Register.
33
®
•
Other registers that should restore only certain bits from the saved-state values:
Bit Blit Control
MM 0x7000c
•
Read the current value of the Bit Blit Control Register.
•
Clear the bits pertaining to the Color Expansion Mode (bits 5:4).
•
OR–in the saved value of the Bit Blit Control Register.
•
Write the result back into the Bit Blit Control Register.
Display Control Register
MM 0x70008
•
Read the current value of the Display Control Register.
•
OR–in the saved value of the Display Control Register.
•
Write the result back into the Display Control Register.
MM 0x70009
•
Read the current value of the Pixel Pipeline Configuration 0 Register.
•
Save reserved bits 6:5 and 2. Clear all other bits.
•
OR–in the saved value of the Pixel Pipeline Configuration 0 Register.
•
Write the result back into the Pixel Pipeline Configuration 0 Register.
MM 0x7000b
•
•
Save reserved bits 7:4 and 1:0. Clear all other bits.
•
•
34
R
MM 0x7000a
•
•
Clear the Display Color Mode bit (bits 3:0).
•
•
®
R
MM 0x2098
•
Read the current value of the Hardware Status Mask Register.
•
Clear everything but the reserved bits (14:13).
•
OR–in the saved value of the Hardware Status Mask Register.
•
Write the result back into the Hardware Status Mask Register.
MM 0x20a0
•
Read the current value of the Interrupt Enable Register.
•
•
OR–in the saved value of the Interrupt Enable Register.
•
Write the result back into the Interrupt Enable Register.
MM 0x20a8
•
Read the current value of the Interrupt Mask Register.
•
•
OR–in the saved value of the Interrupt Mask Register.
•
Write the result back into the Interrupt Mask Register.
Error Mask Register
MM 0x20b4
•
Read the current value of the Error Mask Register.
•
•
OR–in the saved value of the Error Mask Register.
•
Write the result back into the Error Mask Register.
Watermark and Burstlength Control
MM 0x20d8
•
Read the current value of the Watermark and Burstlength Control Register.
•
Clear the burst length and watermark bits (bits 22:20, 17:12, 10:8 and 5:0).
•
OR–in the saved value of the Watermark and Burstlength Control Register.
•
Write the result back into the Watermark and Burstlength Control Register.
35
®
•
•
Disable the low-priority ring buffer, in preparation for setting new values, by clearing the
RING_VALID bit in the Low-Priority Ring Buffer Length field at MM 0x203c.
•
Read the current value of the Low-Priority Ring Buffer Length field (MM 0x203c).
•
Clear the valid bit (bit 0).
•
Write the result back into the Low-Priority Ring Buffer Length field.
Set up the low-priority ring buffer.
•
Write a 0 to the low-priority ring buffer tail at MM 0x2030.
•
Write a 0 to the low-priority ring buffer head at MM 0x2034.
•
Restore the low-priority ring buffer start at MM 0x2038, but preserve the reserved bits.
•
Restore the Low-Priority Ring Buffer Length field, but preserve the Automatic Report Header
Pointer bits and set the Ring Buffer Valid flag.
•
Turn on the screen.
•
Relock the protected register space in order to complete the state restoration process.
At this point the graphics adapter should function completely, in the mode identified by the saved-state information.
36
R
®
R
6. BLT Engine Programming
6.1
6.1.1
BLT Engine Programming Considerations
When the Source and Destination Locations Overlap
It is possible to have BLT operations in which the locations of the source and destination data overlap. This frequently
occurs in BLT operations where a user is shifting the position of a graphical item on the display by only a few pixels. In
these situations, the BLT engine must be programmed so that destination data is not written into destination locations
that overlap source locations, before the source data at those locations has been read. Otherwise, the source data will
become corrupted.
The following figure shows how the source data can be corrupted when a rectangular block is copied from a source
location to an overlapping destination location. The BLT engine reads from the source location and writes to the
destination location, starting with the leftmost pixel in the top line of both, as shown in Step (a). As shown in Step (b),
corruption of the source data already started with the copying of the top line in Step (a). Part of the source that originally
contained lighter pixels now has been overwritten with darker pixels. More source data corruption occurs as Steps (b)
through (d) are performed. At Step (e), another line of source data is read, but the two rightmost pixels of this line are in
the region where the source and destination locations overlap and where the source has already been overwritten as a
result of the copying of the top line in Step (a). Starting in Step (f), darker pixels can be seen in the destination where
lighter pixels should be. This errant effect occurs repeatedly during the remaining steps of this BLT operation. As more
lines are copied from the source location to the destination location, it becomes clear that the end result is not what was
intended originally.
37
®
R
(b)
Source
(c)
(a)
Destination
(d)
(e)
(f)
Source
(g)
(i)
Destination
(h)
b_blt2.vsd
Figure 12. Source corruption in BLT with overlapping source and destination locations
38
®
R
When necessary, the BLT engine can alter the order in which source data is read and destination data is written, in order
to avoid source data corruption problems when the source and destination locations overlap. The command packets
provide the ability to change the point at which the BLT engine begins reading and writing data from the upper-left-hand
corner (the usual starting point) to one of the other three corners. The BLT engine may be set to read data from the
source and write it to the destination, starting at any of the four corners of the panel.
(b)
Source
(c)
(a)
Destination
(d)
(e)
(f)
Source
(g)
(i)
Destination
(h)
b_blt3.vsd
Figure 13. Correctly performed BLT with overlapping source and destination locations
39
®
R
The figure below illustrates how this BLT engine feature can be used to perform the BLT operation illustrated in the
figure above, while avoiding source data corruption. As shown in the figure below, the BLT engine reads the source data
and writes the data to the destination, starting with the right-most pixel of the bottom line. In this manner, no pixel at the
overlap of the source and destination locations ever will be written to before it is read from by the BLT engine. By the
time the BLT operation has reached Step (e), where two pixels at the overlap of the source and destination locations are
about to be overwritten, the source data for those two pixels has already been read.
Destination
Destination
Destination
Destination
OR
Source
Destination
Source
Source
Source
Source
Source
OR
Destination
Destination
OR
Source
Source
Source
Source
Source
Destination
Source
OR
Destination
Destination
Destination
Destination
b_blt4.vsd
Figure 14. Suggested starting points for possible source and destination overlaps
The figure above shows the recommended lines and pixels to be used as starting points in each of the eight possible ways
in which the source and destination locations may overlap. In general, the starting point should be within the area of
source and destination overlap.
40
®
R
6.2
6.2.1
Basic Graphics Data Considerations
Contiguous vs. Discontinuous Graphics Data
Graphics data stored in memory, particularly in the frame buffer of a graphics system, has organizational characteristics
that often distinguish them from other varieties of data. The main distinction is the tendency for graphics data to be
organized in a discontinuous block made up of multiple sub-blocks of bytes, instead of a single contiguous block of
bytes.
(0, 0)
(639, 0)
63
32 31
0
270F8h
28100h
28108h
256, 256
261, 256
256th Scan Line
Note: Drawing is not to scale
(0, 479)
(639, 479)
b_blt5.vsd
Figure 15. Representation of on-screen single 6-pixel line in the frame buffer
The figure above shows an example of contiguous graphics data: a horizontal line made up of six adjacent pixels within a
single scan line on a display with a resolution of 640×480. Presuming that the graphics system driving this display has
been set to 8 bits per pixel and that the frame buffer’s starting address (0h) corresponds to the upper-leftmost pixel of this
display, the six pixels that make this horizontal line starting at coordinates (256, 256) would occupy six bytes, starting at
frame buffer address 28100h and ending at address 28105h.
In this case, there is only one scan line’s worth of graphics data in this single horizontal line, so the block of graphics
data for all six of these pixels exists as a single, contiguous block comprised only of these six bytes. The starting address
and the number of bytes are the only pieces of information that a BLT engine would require to read this block of data.
The simplicity of the preceding example of a single horizontal line contrasts sharply with the example of discontinuous
graphics data depicted in the following figure. The simple six-pixel line of the figure above now is accompanied by three
more six-pixel lines placed on subsequent scan lines, resulting in the 6×4 block of pixels shown.
41
®
R
(0, 0)
(639, 0)
63
32 31
0
270F8h
28100h
28108h
256, 256
261, 256
270F8h
28100h
28108h
256th Scan Line
257th Scan Line
258th Scan Line
259th Scan Line
256, 259
270F8h
28100h
28108h
261, 259
270F8h
28100h
28108h
(0, 479)
(639, 479)
b_blt6.vsd
Figure 16. Representation of on-screen 6×
×4 array of pixels in the frame buffer
On each of the scan lines in which this 6×4 block exists, there are other pixels that are not part of this 6×4 block, so what
appears to be a single 6×4 block of pixels on the display must be represented by a discontinuous block of graphics data
made up of four separate sub-blocks of six bytes apiece, in the frame buffer at addresses 28100h, 28380h, 28600h, and
28880h. This situation makes more complex the task of reading what appears to be a simple 6×4 block of pixels.
However, there are two characteristics of this 6×4 block of pixels that help simplify the task of specifying the locations
of all 24 bytes of this discontinuous block of graphics data: all four of the sub-blocks are of the same length and the four
sub-blocks are separated from each other at equal intervals.
The BLT engine is designed to make use of these characteristics of graphics data, in order to simplify the programming
required to handle discontinuous blocks of graphics data. For such a situation, the BLT engine requires only four pieces
of information: the starting address of the first sub-block, the length of a sub-block, the offset (in bytes) (i.e., the pitch) of
the starting address of each subsequent sub-block, and the quantity of sub-blocks.
6.2.2
Source Data
The source data may exist in the frame buffer or main memory graphics memory, where the BLT engine may read it
directly, or it may be provided to the BLT engine by the host CPU through the command packets. The block of source
graphics data may be either contiguous or discontinuous and may be either in color (with a color depth that matches that
to which the BLT engine has been set) or monochrome.
The source select bit in the command packets specifies whether the source data exists in the frame buffer or are provided
through the command packets. Monochrome source data always is specified as being supplied through an immediate
command packet.
If the color source data resides within the frame buffer or main memory’s graphics memory, then the Source Address
Register specified in the command packets is used to specify the address of the source. However, if the host CPU
42
®
R
provides the source data, then this register takes on a different function, and the three least-significant bits of the Source
Address Register can be used to specify the number of bytes that must be skipped in the first quadword received from the
command packet, in order to reach the first byte of valid source data.
In cases where the host CPU provides the source data, it does so by writing the source data to the ring buffer directly
after the BLT command that requires the data or uses an IMMEDIATE_INDIRECT_BLT command packet that has a
size and pointer to the operand in main memory’s graphics memory.
There also is an address space used for debug, where the CPU can write the source data. It is a 64-KB memory space on
the host bus. There is no actual memory allocated to this memory space, so any data that is written to this location cannot
be read back. This memory space is simply a range of memory addresses that the BLT engine’s address decoder watches
for the occurrence of any memory writes.
The BLT engine loads all data written to any memory address within this memory space or through the command packet,
in the order in which they are written, regardless of the specific memory address to which they are written, and it then
uses that data as the source data in the current BLT operation. The block of bytes sent by the host CPU to either this data
port or through the command packets must be quadword-aligned, although the source data contained within the block of
bytes does not need to be aligned. As mentioned previously, the least-significant three bits of the Source Address
Register are used to specify the number of bytes that must be skipped in the first quadword of color data, in order to
reach the first byte of valid source data.
To accommodate discontinuous source data, the source and destination pitch registers can be used to specify the offset,
in bytes, from the beginning of one scan line’s worth of source data to the next. Otherwise, if the source data is
contiguous, then an offset equal to the length of a scan line’s worth of source data should be specified.
6.2.3
Monochrome Source Data
The opcode of the command packet specifies whether the source data is color or monochrome. Since monochrome
graphics data only uses one bit per pixel, each byte of monochrome source data typically carries data for eight pixels,
which hinders the use of byte-oriented parameters when specifying the location and size of valid source data.
Monochrome source data always is supplied through the command stream, which avoids the read latency during BLT
engine operation. Some additional parameters must be specified in order to ensure the proper reading and use of
monochrome source data by the BLT engine. The BLT engine also provides additional options for the manipulation of
monochrome source data versus color source data.
Various bitwise logical operations and per-pixel write-masking operations were designed to work with color data. In
order to use monochrome data, the BLT engine converts it into color data through a process called color expansion,
which takes place as a BLT operation is performed. In color expansion, the single bits of monochrome source data are
converted into one, two, three or four bytes (depending on the color depth to which the BLT engine has been set) of color
data that are set to carry value corresponding to either the foreground or background color specified for use in this
conversion process. If a given bit of monochrome source data carries the value 1, then the byte(s) of color data resulting
from the conversion process will be set to carry the value of the foreground color. If a given bit of monochrome source
data carries the value 0, then the resulting byte(s) will be set to the value of the background color. The foreground and
background colors used in the color expansion of monochrome source data can be set in the source expansion foreground
color register and the source expansion background color register.
The BLT engine requires the specification of the bit alignment of each scan line’s worth of monochrome source data.
Each scan line’s worth of monochrome source data is word-aligned, but it actually can start on any bit boundary of the
first byte. Monochrome text is a special case and is bit-packed, such that there are no invalid pixels (bits) between scan
lines. Mono Source Start is a three-bit field that indicates the starting pixel position within the first byte for each scan
line.
43
®
R
The BLT engine also provides various clipping options for use with specific BLT commands (BLT_TEXT), with a
monochrome source. Clipping is supported via clip rectangle Y addresses and X coordinates, scan line starting and
ending addresses, and X starting and ending coordinates.
6.2.4
Pattern Data
The color pattern data must exist within the frame buffer or main memory’s graphics memory, where the BLT engine
may read it directly. Monochrome pattern data is supplied by the command packet when it is to be used. As shown in
following figure, the block of pattern graphics data always is represented as an 8×8 block of pixels. The bits or bytes of a
block of pattern data may be organized in the frame buffer memory in only one of four ways, depending upon its color
depth, which may be 8, 16, 24 or 32 bits per pixel (whichever matches the color depth to which the BLT engine has been
set) or monochrome.
Pixel (0, 0)
Pixel (7, 0)
Pixel (0, 7)
Pixel (7, 7)
b_blt7.vsd
63
Pixel
(0, 7)
57 56
48 47
40 39
32 31
Pixel
(7, 7)
24 23
Pixel
(0, 0)
16 15
8 7
0
Pixel
(7, 0)
b_blt8.vsd
Figure 17. Pattern data (always an 8×
×8 array of pixels)
The Pattern Address Register is used to specify the color pattern data address at which the block of pattern data begins.
The three least-significant bits of the address written to this register are ignored, because the address must be specified in
units of quadwords. This is because the pattern always must be located on an address boundary equal to its size.
Monochrome patterns take up 8 bytes (i.e., a quadword of space) and are loaded through the command packet that uses
it. Similarly, color patterns with color depths of 8, 16, and 32 bits per pixel must start on 64-byte, 128-byte, and 256-byte
boundaries, respectively. Color patterns with color depths of 24 bits per pixel must start on 256-byte boundaries, despite
the fact that the actual color data fills only 3 bytes per pixel. The next four figures show how monochrome, 8-bpp, 16bpp, 24-bpp, and 32-bpp pattern data, respectively, is organized in memory.
44
®
R
63
57 56
48 47
40 39
32 31
24 23
16 15
Pixel (0, 7)
8 7
0
Pixel (0, 0)
00h
08h
10h
18h
20h
28h
30h
Pixel (7, 7)
Pixel (7, 0)
38h
b_blt9.vsd
Figure 18. 8-bpp pattern data — Occupies 64 bytes (8 quadwords)
63
48 47
32 31
16 15
0
Pixel (0, 0)
Pixel (7, 0)
00h
08h
68h
Pixel (0, 7)
Pixel (7, 7)
70h
78h
b_blt10.vsd
45
®
63
56 55
R
48 47
40 39
32 31
16 15
24 23
Pixel (0, 0)
Green
Red
0
8 7
00h
Blue
08h
Red
Pixel (7, 0)
Green
10h
Blue
"Throw-Away" Fourth Bytes for Pixels (0, 0) through (7, 0)
18h
D8h
Pixel (0, 7)
Green
Red
E0h
Blue
E8h
Red
Pixel (7, 7)
Green
Blue
F0h
"Throw-Away" Fourth Bytes for Pixels (0, 7) through (7, 7)
F8h
b_blt11.vsd
63
48 47
32 31
16 15
Pixel (0, 0)
Pixel (3, 0)
0
00h
08h
68h
Pixel (4, 7)
Pixel (7, 7)
70h
78h
b_blt10.vsd
As is shown in the 24-bpp pattern data figure, four bytes are allocated for each pixel on each scan line’s worth of pattern
data, which allows each scan line’s worth of 24-bpp pattern data to begin on a 32-byte boundary. The extra (i.e.,
“fourth”) unused bytes of each pixel on a scan line’s worth of pattern data are collected together in the last 8 bytes (i.e.,
the last quadword) of each scan line’s worth of pattern data.
The opcode of the command packet specifies whether the pattern data is color or monochrome. The various bitwise
logical operations and per-pixel write-masking operations were designed to work with color data. In order to use
monochrome pattern data, the BLT engine is designed to convert it to color, through a process called “color expansion”
that takes place as a BLT operation is performed. In color expansion, the single bits of monochrome pattern data are
converted to one, two, three or four bytes (depending on the color depth to which the BLT engine has been set) of color
data that are set to carry values corresponding to either the foreground or background color specified for use in this
process. The foreground color is used for pixels corresponding to a bit of monochrome pattern data that carries the value
1, while the background color is used where the corresponding bit of monochrome pattern data carries the value 0. The
foreground and background colors used in the color expansion of monochrome pattern data can be set in the Pattern
Expansion Foreground Color Register and Pattern Expansion Background Color Register.
46
®
R
6.2.5
Destination Data
There are actually two different types of “destination data”: the graphics data already residing at the location designated
as the destination and the data to be written to that location, as a result of a BLT operation.
The location designated as the destination must be within the frame buffer or the main memory’s graphics memory,
where the BLT engine can directly read from it and write to it. The blocks of destination data to be read from and written
to the destination may be either contiguous or discontinuous. All data written to the destination will have the color depth
to which the BLT engine has been set. It is presumed that any data already existing at the destination to be read by the
BLT engine also will be of this same color depth. The BLT engine neither reads nor writes monochrome destination
data.
The Destination Address Register is used to specify the address of the destination.
To accommodate discontinuous destination data, the Source and Destination Pitch Registers can be used to specify the
offset, in bytes, from the beginning of one scan line’s worth of destination data to the next. Otherwise, if the destination
data is contiguous, then an offset equal to the length of a scan line’s worth of destination data should be specified.
47
®
6.3
R
BLT Programming Examples
6.3.1
Pattern Fill — A Very Simple BLT
In this example, a rectangular area on the screen will be filled with a color pattern stored as pattern data in off-screen
memory. The screen has a resolution of 1024×768, and the graphics system has been set to a color depth of 8 bits per
pixel.
63
(0, 0)
128, 128
0
(1023, 0)
128, 128
Rectangular
Area to be Filled
(Destination)
191, 128
Scan Lines 128 Through 191
20080h
20088h
20090h
20098h
200A0h
200A8h
200B0h
200B8h
On
128th
Scan
Line
2FC80h
2FC88h
2FC90h
2FC98h
2FCA0h
2FCA8h
2FCB0h
2FCB8h
On
191th
Scan
Line
(191, 128)
128, 191
191, 191
(128, 191)
(0, 767)
(1023, 767)
(191, 191)
b_blt20.vsd
Figure 22. On-screen destination for example pattern fill BLT
As shown in the preceding figure, the upper left-hand corner of the rectangular area to be filled is at coordinates (128,
128), while its lower right-hand corner is at coordinates (191, 191). These coordinates define a rectangle covering 64
scan lines, each line of which is 64 pixels in length (i.e., an array of 64×64 pixels). Assuming that the pixel at
coordinates (0, 0) corresponds to the byte at address 00h in the frame buffer memory, the pixel at (128, 128) corresponds
to the byte at address 20080h.
48
®
R
63
Pattern Data
(0, 0)
(7, 0)
(0, 7)
(7, 7)
(7, 0)
(0, 0) 0
100000h
100008h
100010h
100018h
100020h
100028h
100030h
100038h
(7, 7)
(0, 7)
b_blt22.vsd
Figure 23. Pattern data for example pattern fill BLT
As shown in the preceding figure, the pattern data occupies 64 bytes, starting at address 100000h. As always, the pattern
data represents an 8×8 array of pixels.
The BLT command packet is used to select the features to be used in this BLT operation, so it must be programmed
carefully. The vertical alignment field should be set to 0, in order to select the top horizontal row of the pattern as the
starting row used in drawing the pattern, starting with the top scan line covered by the destination. The pattern data is in
color with a color depth of 8 bits per pixel, so dynamic color enable should be asserted with the dynamic color depth
field set to 0. Since this BLT operation does not use per-pixel write-masking (i.e., the destination transparency mode),
this field should be set to 0. Finally, the raster operation field should be programmed with the 8-bit value F0h, in order to
select the bitwise logical operation in which a simple copy of the pattern data to the destination takes place. Selecting
this bitwise operation, in which no source data is used as an input, causes the BLT engine to automatically forego either
reading source data from the frame buffer or waiting for the host CPU to provide it.
The Destination Pitch Register must be programmed with number of bytes in the interval from the start of one scan line’s
worth of destination data to the next. Since the color depth is 8 bits per pixel and the horizontal resolution of the display
is 1024, the value to be programmed into these bits is 400h, which is equal to the decimal value 1024.
Bits [31:3] of the Pattern Address Register must be programmed with the address of the pattern data.
Similarly, bits [31:0] of the Destination Address Register must be programmed with the byte address at the destination
that will be written to first. In this case, the address is 20080h, which corresponds to the byte representing the pixel at
coordinates (128, 128).
This BLT operation does not use the values in the Source Address Register or the Source Expansion Background or
Foreground Color Registers.
The Destination Width and Height Registers must be programmed with values that tell the BLT engine the 64×64 pixel
size of the destination location. The height should be set to carry the value 40h, indicating that the destination location
covers 64 scan lines. The width should be set to carry the value 40h, indicating that each scan line’s worth of destination
data occupies 64 bytes. All of this information is written to the ring buffer using the PAT_BLT command packet.
49
®
R
63
(0, 0)
128, 128
0
(1023, 0)
128, 128
191, 128
20080h
20088h
20090h
20098h
200A0h
200A8h
200B0h
200B8h
On
128th
Scan
Line
2FC80h
2FC88h
2FC90h
2FC98h
2FCA0h
2FCA8h
2FCB0h
2FCB8h
On
191th
Scan
Line
(191, 128)
(128, 191)
128, 191
191, 191
(0, 767)
(1023, 767)
(191, 191)
b_blt21.vsd
Figure 24. Results of example pattern fill BLT
The figure above shows the end result of performing this BLT operation. The 8×8 pattern has been repeatedly copied
(“tiled”) into the entire 64×64 area at the destination.
50
®
R
6.3.2
Drawing Characters Using a Font Stored in System Memory
In this example BLT operation, a lowercase “f” is to be drawn in black on a display with a gray background. The
resolution of the display is 1024×768, and the graphics system has been set to a color depth of 8 bits per pixel.
(0, 0)
(1023, 0)
128, 128
63
0
20080h
128th Scan Line
20480h
129th Scan Line
20880h
130th Scan Line
128, 128
20C80h
131th Scan Line
21080h
132nd Scan Line
21480h
133rd Scan Line
Destination
21880h
134th Scan Line
21C80h
135th Scan Line
(135, 135)
135, 135
(0, 767)
(1023, 767)
b_blt12.vsd
Figure 25. On-screen destination for example character drawing BLT
The preceding figure shows the display on which this “f” is to be drawn. As shown in this figure, the entire display has
been filled with gray. The “f” is to be drawn in an 8×8 region on the display, with the upper-left-hand corner at the
coordinates (128, 128).
Source Data
Pixel (0, 0)
Pixel (7, 0)
63
57 56
48 47
40 39
32 31
24 23
16 15
8 7
0
00000000 00010000 00010000 00111100 00010000 00010000 00001100 00000000
(0, 7)
Pixel (0, 7)
(7, 7)
(7, 0)
(0, 0)
Pixel (7, 7)
b_blt13.vsd
Figure 26. Source data in system memory for example character drawing BLT
51
®
R
The figure above shows both the 8×8 pattern making up the “f” and how it is represented somewhere in the host’s system
memory. (The actual address in system memory is not important.) The “f” is represented in system memory by a block of
monochrome graphics data that occupies 8 bytes. Each byte carries the 8 bits needed to represent the 8 pixels in each
scan line’s worth of this graphics data. This type of pattern often is used to store character fonts in system memory.
During this BLT operation, the host CPU will read this representation of “f” from system memory and will write it to the
BLT engine, by performing memory writes to the ring buffer as an immediate monochrome BLT operand following the
BLT_TEXT command. The BLT engine will receive this data through the command stream and will use it as the source
data for this BLT operation. The BLT engine will be set to the same color depth as the graphics system (in this case, 8
bits per pixel). Since the source data in this BLT operation is monochrome, color expansion must be used to convert it to
an 8-bpp color depth. To ensure that the gray background behind this “f” is preserved, per-pixel write-masking will be
performed, using the monochrome source data as the pixel mask.
The BLT Setup and Text command packets are used to select the features to be used in this BLT operation. Only the
fields required by these two command packets must be programmed carefully. The BLT engine ignores all other registers
and fields. The source select field must be set to 1, in order to indicate that the source data is provided by the host CPU
through the IMMEDIATE_BLT command packet. Finally, the raster operation field should be programmed with the 8bit value CCh, in order to select the bitwise logical operation that simply copies the source data to the destination.
Selecting this bitwise operation, in which no pattern data is used as an input, causes the BLT engine to automatically
forego reading pattern data from the frame buffer.
The Setup Pattern/Source Expansion Foreground Color Register is used to specify the color with which the “f” will be
drawn. There is no source address. All scan lines of the glyph are bit-packed, and the clipping is controlled by the
ClipRect registers from the SETUP_BLT command and the Destination Y1, Y2, X1, and X2 registers in the TEXT_BLT
command. Only the pixels that are within the clip rectangle are written to the destination surface.
The Destination Pitch Register must be programmed with a value equal to the number of bytes in the interval between
the first bytes of each adjacent scan line’s worth of destination data. Since the color depth is 8 bits per pixel and the
horizontal resolution of the display is 1024 pixels, the value to be programmed into these bits is 400h, which is equal to
the decimal value 1024. Since the source data used in this BLT operation is monochrome, the BLT engine will not use a
byte-oriented pitch value for the source data.
Since the source data is monochrome, color expansion is required to convert it to color with a color depth of 8 bits per
pixel. Since the Setup Pattern/Source Expansion Foreground Color Register is selected in order to specify the foreground
color of black to be used in drawing the “f”, this register must be programmed with the value for that color. With the
graphics system set for a color depth of 8 bits per pixel, the actual colors are specified in the RAMDAC palette, and the 8
bits stored in the frame buffer for each pixel actually specify the index used to select a color from that palette. This
example assumes that the color specified at index 00h in the palette is black, and therefore bits [7:0] of this register
should be set to 00h in order to select black as the foreground color. The BLT engine ignores bits [23:8] of this register
because the selected color depth is 8 bits per pixel. Even though the color expansion being performed on the source data
normally requires the specification of both the foreground and background colors, the value used to specify the
background color is not important in this example. Per-pixel write-masking is being performed with the monochrome
source data as the pixel mask, which means that none of the pixels in the source data converted to the background color
will ever be written to the destination. Since these pixels will never be seen, the value programmed into the
Pattern/Source Expansion Background Color Register in order to specify a background color is not important.
The Destination Width and Height Registers are not used. The Y1, Y2, X1, and X2 registers are programmed with values
that describe to the BLT engine the 8x8 pixel size of the destination location. The Destination Y1 and Y2 address
registers must be programmed with the starting and ending scan line addresses of the destination data. These addresses
are specified as an offset from the start of the frame buffer of the scan line at the destination that will be written to first.
The destination X1 and X2 registers must be programmed with the starting and ending pixel offsets from the beginning
of the scan line.
52
®
R
This BLT operation does not use the values in the Pattern Address Register, the Source Expansion Background Color
Register, or the Source Expansion Foreground Color Register.
(0, 0)
(1023, 0)
128, 128
63
0
20080h
(128th Scan Line)
20480h
(129th Scan Line)
20880h
(130th Scan Line)
128, 128
20C80h
(131th Scan Line)
21080h
(132nd Scan Line)
21480h
(133rd Scan Line)
Destination
21880h
(134th Scan Line)
21C80h
(135th Scan Line)
135, 135
135, 135
(0, 767)
(1023, 767)
b_blt12.vsd
Figure 27. Results of example character-drawing BLT
Figure 27 shows the end result of performing this BLT operation. Only the pixels that form part of the actual “f” have
been drawn in the 8×8 destination location on the display, leaving the other pixels within the destination with their
original gray color.
53
®
R
54
®
R
7. Initialization Registers
In order to function, all registers described in this section must be programmed for the Intel® 82180 chipset and family
products. The default states of these registers, with the exception of registers that deal with extended modes or
performance enhancements, will prevent the Intel 82180 chipset and family products from booting.
Note: The registers in this document are normally programmed by the video BIOS.
These registers also may be documented in other sections of this document.
7.1
Standard VGA Registers
All VGA registers are in standard locations and initialized by means of standard procedures. This section will document
all nonstandard registers that are needed for initialization of the Intel 82180 chipset and family chipsets.
7.2
SMRAM Registers
7.2.1
SMRAM—System Management RAM Control Register (Device 0)
Address offset:
Default value:
Access:
Size:
70h
00h
Read/Write
8 bits
The SMRAM register controls how accesses to compatible and extended SMRAM spaces are treated as well as how
much (if any) memory is “taken” from the system in order to support both SMRAM and graphics local memory needs.
7
6
Graphics Mode Select
5
4
Upper SMM Select
3
2
Lower SMM Select
1
0
SMM
Space
Locked
E_SMRA
M_ERR
55
®
Bit
7:6
R
Description
Graphics Mode Select (GMS). This field is used to enable/disable the internal graphics device and
select the amount of Main Memory that is “taken” to support the internal graphics device.
00 = Graphics device disabled, no memory “taken” (Device 1 is NOT accessible in this case.)
01 = Graphics device enabled, no memory “taken”
10 = Graphics device enabled, 512 KB of memory “taken”
11 = Graphics device enabled, 1 MB of memory “taken”
Note:
When the graphics device is disabled (00), the graphics device and all of its memory and I/O functions
are disabled, the clocks to this logic are turned off, memory accesses to the VGA range (A0000BFFFF) will be forwarded to the hub-link, and the graphics local memory space is NOT “taken” from
main memory. When this field is non-0, the GMCH graphics device and all of its memory and I/O
functions are enabled, all non-SMM memory accesses to the VGA range will be handled internally,
and the selected amount of graphics local memory space (0, 512K or 1M) is “taken” from the main
memory. Graphics memory is “taken” AFTER TSEG memory is “taken”.
Once D_LCK is set, these bits becomes Read Only.
GMCH does not support VGA on local memory. Software must not use the 01 mode for VGA.
5:4
Upper SMM Select (USMM). This field is used to enable/disable the various SMM memory ranges
above 1 Meg. TSEG is a block of memory (“taken” from main memory at [TOM-Size] : [TOM]) that is
accessible only by the processor and only while operating in the SMM mode. HSEG is a remap of the
AB segment at FEEA0000 : FEEBFFFF. Both of these areas, when enabled, are usable as SMM RAM.
Non-SMM operations that use these address ranges are forwarded to the hub-link. HSEG is ONLY
enabled if LSMM = 00.
00 = TSEG and HSEG are both disabled.
01 = TSEG is disabled and HSEG is conditionally enabled.
10 = TSEG is enabled as 512 Kbytes and HSEG is conditionally enabled.
11 = TSEG is enabled as 1 Mbytes and HSEG is conditionally enabled.
Once D_LCK is set, these bits become Read Only.
3:2
Lower SMM Select (LSMM). This field controls the definition of the A&B segment SMM space.
00 = AB segment disabled
01 = AB segment enabled as general system RAM
10 = AB segment enabled as SMM code RAM shadow. Only SMM code reads can access DRAM in the
AB segment. SMM data operations and all non-SMM operations either go to the internal graphics
device or are broadcast on the hub-link.
11 = AB segment enabled as SMM RAM. All SMM operations to the AB segment are serviced by
DRAM. All non-SMM operations either go to the internal graphics device or are broadcast on hublink.
When D_LCK is set, bit 3 becomes Read_Only and bit 2 is writeable ONLY if bit 3 is 1.
1
SMM Space Locked (D_LCK): When D_LCK is set to 1, then D_LCK, GMS, USMM, and the mostsignificant bit of LSMM become Read Only. D_LCK can be set to 1 via a normal configuration space
write but can be cleared only by a reset. The combination of D_LCK and LSMM provides convenience
with security. The BIOS can use LSMM=01 to initialize SMM space and then use D_LCK to “lock down”
SMM space in the future, so that no application software (or the BIOS itself) can violate the integrity of
SMM space, even if the program has knowledge of the LSMM function. This bit also locks the DRP
register.
0
E_SMRAM_ERR (E_SMERR): This bit is set when CPU accesses the defined memory ranges in
extended SMRAM (HSEG or TSEG) while not in SMM mode. It is software’s responsibility to clear this
bit. The software must write a 1 to this bit to clear it This bit is NOT set in the case of an explicit writeback operation.
56
®
R
Initialization and Usage of “Taken” Memory
SMRAM Register bits 7:4 control the theft of memory from main memory space, for use as graphics local memory and
SMM TSEG memory. The blocks of memory selected by these fields are NOT accessible as general system RAM. When
bit 5 of the SMRAM register is 1, the TSEG segment of memory can ONLY be accessed by the CPU in the SMM mode.
(No other agent can access this memory.) Therefore, the BIOS should initialize this block of memory BEFORE setting
either bit 5 or bit 7 of the SMRAM register. The memory for TSEG is “taken” first, and then the graphics local memory
is “taken.” An example of this theft mechanism is as follows:
TOM = 64 Mbytes.
TSEG selected as 512 Kbytes in size.
Graphics local memory selected as 1 Mbyte in size.
General System RAM available in system = 62.5 Mbytes.
General system RAM range:
00000000h to 03E7FFFFh
TSEG address range:
03F80000h to 03FFFFFFh
TSEG “taken” from:
03F80000h to 03FFFFFFh
Graphics local memory “taken” from: 03E80000h to 03F7FFFFh
7.3
Graphics Controller Registers
The graphics controller registers are accessed via either I/O space or memory space. To access the registers, the VGA
Graphics Controller Index Register at I/O address 3CEh (or memory address 3CEh) is written with the index of the
desired register, and then the desired register is accessed through the data port for the graphics controller registers at I/O
address 3CFh (or memory address 3CFh).
7.3.1
GR10
Address Mapping
I/O (and memory offset) address:
Default:
Attributes:
7
3CFh (Index=10h)
00h
R/W
5
Reserved (0000)
4
3
2
1
0
Paging to
LM
VGA
Buffer
/Memory
Map
Packed
Mode
Enabling
Linear
Mapping
Page
Mapping
57
®
Bit
7:5
4
R
Description
Reserved (000)
Page to Local Memory Enable:
Used only if GR10(0) = 1 {paging enabled} and (GR10(1) = 1 or GR10(2) = 1) {either packed mode or
linear mode is enabled) and GR10(3) = 0 {VGA buffer selected}.
0 = Page to VGA buffer
1 = Page to physical local memory
3
VGA Buffer/Memory Map Select.
0 = VGA buffer (default)
1 = Memory map
2
Packed Mode Enable.
0 = Address and data translation are bused register settings (default).
1 = Forced extended pack pixel address translation
In the page mapping mode, register GR06 selects the video memory address.
1
Linear Mapping (PCI).
0 = Disable (default)
1 = Enable
0
Page Mapping Enable. This mode allows the mapping of the VGA space allocated in main memory (nonlocal video memory) mode or all of local memory space through the [A0000:AFFFF] window (using bit 4 of
this register), which is a 64-KB page. An internal address is generated using GR11[6:0] as the address
line [22:16] extension to A[15:2].
1 = Enable
58
®
R
Table 3. VGA Address Range
GR10
[2]
GR10
[1]
GR10
[0]
0
0
0
0
0
1
0
1
0
0
1
1
1
0
0
1
0
1
1
1
0
1
1
1
Note 1
Std VGA
xlations
Paging and
VGA xlation
No paging, no
VGA xlations
Paging, no
VGA xlations
No paging, no
VGA xlations
Paging, no
VGA xlations
No paging, no
VGA xlations
Paging, no
VGA xlations
Address Range (see Note 2)
A0000-AFFFF Range
(No GTT)
VGA controller, no paging
B0, B8 Ranges
(No GTT)
VGA controller, paged by GR11
Bypass VGA, no paging
Bypass VGA, paged by GR11
Bypass VGA paged by GR11
Notes:
GR10[2:0]: 001 should not be used for paging, because all the VGA registers need to be setup correctly. An access
thru A0000 range is paged by GR11. Note that prefetch refers to the cache line size access to GM vs. without
prefetch (i.e., QW).
VGA Address Range is selected by GR06. Graphics range is selected through the graphics base address register in
the configuration space. Access to the VGA range does not require a translation table. VGA range paging
allows access to all of local memory, if it is set up with bit 4 of this register, or to all of the memory taken from
the VGA main memory space. Access to the graphics range requires GTT to be set up and will result in a
prefetch unless prefetch is disabled. Access to VGA range will not result in a prefetch.
The BIOS should access local memory through the "back door" mechanism, by setting gr10 = 17h, gr11 = 0, and gr6
= 0, only when local memory has been enabled (MMADR+3000h). Otherwise, the system will hang in a snoop
stall forever.
59
®
7.3.2
GR11
Page Selector
Default :
Attributes:
Bit
7:0
60
R
3CFh (Index=11h)
00h
R/W
Description
Page Select. Selects a 64-KB window within VGA space in NLVM mode or all of local memory when
page mapping is enabled (GR10[0]=1). In addition, this register is used for page selection of memorymapped register addresses.
®
R
7.4
CRT Controller Register
The CRT controller registers are accessed by writing the index of the desired register into the CRT Controller Index
Register at I/O address 3B4h or 3D4h, depending on whether the graphics system is configured for MDA or CGA
emulation. The desired register then is accessed through the data port for the CRT controller registers located at I/O
address 3B5h or 3D5h, again depending upon the choice of MDA or CGA emulation, as per MSR[0]. For memorymapped accesses, the Index register is at 3B4h (MDA mode) or 3D3h (CGA mode), and the data port is accessed at
3B5h (MDA mode) or 3D5h (CGA mode).
Notes:
Register CR80 enables / disables the CRTC extensions.
Group 0 Protection: In the original IBM VGA, CR[0:7] could write-protected by means of CR11[7]. In the BIOS
code, this write protection is set following each mode change. Other protection groups are not currently used and
will not by used in the future by the BIOS or by drivers. They are the result of an industry fad some years ago,
that attempted to write-protect other groups of registers. However, all such schemes were chip specific. Only
IBM-compatible write protection (Group 0 protection) is supported.
CR30
Extended Vertical Total Register
7.4.1
Default:
Attributes:
3B5h/3D5h (index=30h)
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical total bits 11:8
Description
7:4
Reserved. Read as 0s. This field must be 0s when this register is written.
3:0
Vertical Total Bits [11:8]. The vertical total is a 10-bit or 12-bit value that specifies the total number of
scan lines. This includes the scan lines both inside and outside of the active display area.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the vertical total is
specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits [7:0] of the
Vertical Total Register (CR06), and the 2 most-significant bits are supplied by bits 5 and 0 of the Overflow
Register (CR07). In standard VGA modes, these 4 bits of this register are not used.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the vertical total is specified
with a 12-bit value. The 8 least-significant bits of this value are supplied by bits [7:0] of the Vertical Total
Register (CR06), and the 4 most-significant bits are supplied by these 4 bits of this register.
This 10-bit or 12-bit value should be programmed to be equal to the total number of scan lines, minus 2.
61
®
R
CR31
Extended Vertical Display End Register
7.4.2
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Display End Bits 11:8
Description
7:4
3:0
Vertical Display End Bits [11:8]. The vertical display enable end is a 10-bit or 12-bit value that specifies
the number of the last scan line within the active display area.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the vertical display
enable end is specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits
[7:0] of the Vertical Display Enable End Register (CR12), and the 2 most-significant bits are supplied by
bits 6 and 1 of the Overflow Register (CR07). In standard VGA modes, these 4 bits of this register are not
used.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the vertical display enable
end is specified with a 12-bit value. The 8 least-significant bits of this value are supplied by bits [7:0] of the
Vertical Display Enable End Register (CR12), and the 4 most-significant bits are supplied by these 4 bits of
this register.
This 10-bit or 12-bit value should be programmed to be equal to the number of the last scan line within in
the active display area. Since the active display area always starts on the 0th scan line, this number should
be equal to the total number of scan lines within the active display area, minus 1.
62
®
R
CR32
Extended Vertical Sync Start Register
7.4.3
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Sync Start Bits 11:8
Description
7:4
3:0
Vertical Sync Start Bits [11:8]. The vertical sync start is a 10-bit or 12-bit value that specifies the
beginning of the vertical sync pulse relative to the beginning of the active display area.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the vertical sync start is
Vertical Sync Start Register (CR10), and the 2 most-significant bits are supplied by bits 7 and 2 of the
Overflow Register (CR07). In standard VGA modes, these 4 bits of this register are not used.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the vertical display end is
Vertical Sync Start Register (CR10), and the 4 most-significant bits are supplied by these 4 bits of this
register.
This 10-bit or 12-bit value should be programmed to be equal to the number of scan lines from the
beginning of the active display area to the start of the vertical sync pulse. Since the active display area
always starts on the 0th scan line, this number should be equal to the number of the scan line on which the
vertical sync pulse begins.
63
®
R
CR33
Extended Vertical Blanking Start Register
7.4.4
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Blanking Start Bits 11:8
Description
7:4
3:0
Vertical Blanking Start Bits [11:8]. The vertical blanking start is a 10-bit or 12-bit value that specifies the
beginning of the vertical blanking period relative to the beginning of the active display area.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the vertical blanking
start is specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits [7:0] of the
Vertical Blanking Start Register (CR15), and the most-significant and second-most-significant bits are
supplied by bit 5 of the Maximum Scan Line Register (CR09) and bit 3 of the Overflow Register (CR07),
respectively. In standard VGA modes, these four bits are not used.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the vertical blanking start is
specified with a 12-bit value. The 8 least-significant bits of this value are supplied by bits {7:0] of the
Vertical Blanking Start Register (CR15), and the 4 most-significant bits are supplied by these 4 bits of this
register.
beginning of the active display area to the beginning of the blanking period. Since the active display area
vertical blanking period begins.
64
®
R
CR35
Extended Horizontal Total Time Register
7.4.5
Default:
Attributes:
00h
Read/Write
7
1
Reserved (0000000)
Bit
Reserved. When this register is written to, these bits should be set to 0.
Extended Horizontal Total (MSB that extends CR00).
0
CR39
Extended Horizontal Blank Time Register
7.4.6
Default:
Attributes:
00h
Read/Write
7
1
Reserved (0000000)
Bit
0
Ext Horiz
Total
Description
7:1
7:1
0
0
Ext Horiz
Total
Description
Reserved.
Extended Horizontal Total (MSB that extends CR5[7], CR3[4:0]).
65
®
7.4.7
R
CR40
Extended Start Address Register
Default:
Attributes:
7
6
Start Addr
Enable
Reserved
(0)
00h
Read/Write
5
0
Start Address Bits 23:18
Bit
Description
7
Extended Mode Start Address Enable. This bit is used only in extended modes, where bit 0 of the I/O
Control Register (CR80) is set to 1, in order to signal the graphics controller to update the start address. In
extended modes, the start address is specified with a 30-bit value. These 30 bits, which are provided by
the Start Address Low Register (CR0D), the Start Address High Register (CR0C), the Extended Start
Address High Register (CR42), and bits [5:0] of this register, are double-buffered and synchronized to
VSYNC to ensure that changes occurring on the screen as a result of changes in the start address always
have a smooth or instantaneous appearance. To change the start address in extended modes, all three
registers must be set for the new value, and then this bit of this register must be set to 1. Only if this is
done will the graphics controller update the start address on the next VSYNC. After this update has been
performed, the graphics controller will set bit 7 of this register back to 0.
6
5:0
Start Address Bits [23:18]. This start address is a 16-bit value that specifies the memory address offset
from the beginning of the frame buffer or a 32-bit buffer address at which begins the data to be shown in the
active display area. (Default: 0)
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the start address is
specified with a 16-bit value. The eight bits of the Start Address High Register (CR0C) provide the eight
most-significant bits of this value, while the eight bits of the Start Address Low Register (CR0D) provide
the eight least-significant bits.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the start address is specified
with a 32-bit value. Bits 31:24 of this value are provided by the Extended Start Address High Register
(CR42). Bits 23:18 of this value are provided by bits 5:0 of this register. Bits 17:10 of this value are
provided by the Start Address High Register (CR0C). Bits 9:2 of this value are provided by the Start
Address Low Register (CR0D). Bits 1:0 of this value are always 0 and therefore are not provided. Note
that, in the extended modes, these 32 bits from these four registers are double-buffered and synchronized
to VSYNC to ensure that changes occurring on the screen as a result of changes in the start address
always have a smooth or instantaneous appearance. To change the start address in extended modes, all
four registers must be set for the new value, and then bit 7 of this register must be set to 1. Only if this is
66
®
R
CR41
Extended Offset Register
7.4.8
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Offset Bits 11:8
Description
7:4
3:0
Offset Bits [11:8]. The offset is an 8-bit or 12-bit value describing the number of words or dwords of frame
buffer memory occupied by each horizontal row of characters. Whether this value is interpreted as the
number of words or dwords is determined by the settings of the bits in the Clocking Mode Register (SR01).
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, the offset is described
with an 8-bit value, all bits of which are provided by the Offset Register (CR13).
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, the offset is described with a
12-bit value. The four most-significant bits of this value are provided by bits [3:0] of this register, and the
eight least-significant bits are provided by the Offset Register (CR13).
This 8-bit or 12-bit value should be programmed to be equal to either the number of words or dwords
(depending on the setting of the bits in the Clocking Mode Register, SR01) of frame buffer memory that is
occupied by each horizontal row of characters.
7.4.9
CR42
Extended Start Address High Register
Default:
Attributes:
Bit
7:0
00h
Read/Write
Description
Start Address High Bits [31:24]. This register provides bits [31:24] of the 32-bit buffer address at which
begins the data to be shown in the active display area. (Default: 0)
most-significant bits of this value, while the eight bits of the CR0D register provide the eight leastsignificant bits.
as a 32-bit value. Bits [31:24] of this value are provided by this register. Bits [23:18] of this value are
provided by bits [5:0] of the Extended Start Address Register (CR40). Bits [17:10] of this value are
provided by the Start Address High Register (CR0C). Bits [9:2] of this value are provided by the Start
Address Low Register (CR0D). Bits [1:0] of this value are always 0 and therefore are not provided. It
should be further noted that, in extended modes, the 30 bits from these four registers are double-buffered
and synchronized to VSYNC, in order to ensure that changes occurring on the screen as a result of
changes in the start address always have a smooth or instantaneous appearance. To change the start
address in extended modes, all four registers must be set for the new value, and then bit 7 of the Extended
Start Address Register (CR40) must be set to 1. Only then will the graphics controller update the start
address on the next VSYNC. When the update is done, the graphics controller sets bit 7 of the Extended
Start Address Register (CR40) back to 0.
67
®
7.4.10
R
CR70
Interlace Control Register
Default:
Attributes:
7
00h
Read/Write
6
0
Interlace
Enable
CRT Half-Line Value
Bit
7
Description
Interlace Enable.
0 = Selects non-interlaced CRT output (default).
1 = Selects interlaced CRT output.
6:0
7.4.11
CRT Half-Line Value. When interlaced CRT output has been selected, the value in this register specifies
the position along the length of a scan line at which the half-line vertical sync pulse occurs for the odd
frame. This half-line vertical sync pulse begins at a position between two horizontal sync pulses on the last
scan line, rather than at a position coincident with the beginning of a horizontal sync pulse at the end of a
scan line.
CR80
I/O Control
Default:
Attributes:
3B5h/3D5h(index 80h)
00h
Read/Write
7
2
Reserved (000000)
Bit
7:2
1
1
0
Attr Cntl
Ext Enbl
CRT Cntl
Int Enbl
Description
Reserved.
Attribute Controller Extensions Enable. Controls whether the attribute registers are accessed with both
index and data at 3C0h (strict VGA mode), or whether they are accessed with 3c0h as the index and with
3C1h as data. It is possible that the BIOS software or driver software might not use the non-VGA mode.
Either method should work, but it must be the method the software is using.
0 = Disable (i.e., strictly VGA compatible mode) (default)
1 = Enable attribute controller extensions
Index and Data of the Attribute Controller registers are accessible at 3C0h in standard VGA. When the
Attribute Controller Extensions are enabled, the index and data are accessible at addresses 3C0h and
3C1h, respectively.
0
CRT Controller Interpretation Enable. This bit modifies responses/functionality to/of registers CR30 and
beyond. See CR06, CR07, CR09, CR0C, CR0D, CR10, CR12, CR13, CR15, CR30, CR32, CR33, CR40,
CR41, and CR42.
0 = Registers have strict VGA interpretation (default).
1 = Registers have extended VGA Interpretation (i.e., access to 4-G space).
68
®
R
7.4.12
CR82
Blink Rate Control
Default:
88h is the VGA default because it is more visually appealing (but the standard
VGA default is 83h).
Read/Write
Attributes
The H/W default for this register does not match the VGA compatibility requirements. The BIOS must make sure to set
this register to correct value.
7
6
5
Character Blink Duty
Cycle
Bit
7:6
0
VSync Blink Rate
Description
Character Blink Duty Cycle. (Character blink also is known as attribute blink.)
00 = 50% duty cycle.
01 = 25% duty cycle
10 = 50% duty cycle (power-on default)
5:0
VSync Blink Rate. Controls the cursor blink rate in terms of the number of vsyncs, as follows: The
programmed value must be the (actual value/2) - 1. (Default: 3)
69
®
7.5
R
Display Control Registers
7.5.1
FW_BLC—FIFO Watermark and Burst Length Control
Address Offset :
Default value:
Access:
Size:
020D8h
22 31 73 17h
Read/Write
32 bits
These control values only apply to hi-res modes of operation. VGA modes ignore the settings of these registers in favor
of fixed values.
For the VGA text mode, character buffer fetches are performed without regard to the space available in the FIFO (since
this data is stored in the character buffer, not the FIFO). Character buffer fetches are performed as a single request of 8
qwords. Font data is fetched one qword at a time, and it will begin when the FIFO has room for 8 character font qwords.
VGA graphics modes will perform requests one at a time, as long as there is room for 1 qword in the FIFO.
Note 1: FIFOs refer to ALL FIFOs in the DSI data path. (That is, the total FIFO space available is the sum of the DSI
FIFO depth and the display engine FIFO depth.) Currently, this depth is 48 qwords.
Note 2: The H/W default is an illegal value: These quantities should never be programmed to zeros.
Note 3: The hardware depends on these registers being set properly, since it is possible to set the request length and
watermarks to states that would cause the overflow of the sync FIFO. For example, assume that a watermark is
set to 33 QW and the request length is set to 32 QWs. After the first two requests have been completed, 64 QWs
will have been written into the sync FIFO. During this time, only 16 QWs will be drained out of the FIFO in
order to be written to the display engine FIFO. Since the sync FIFO in the DSI is only 32QWs deep, this will
result in (64-16-32) = 16 QW overflow of the FIFO.
31
28
27
Overlay Delay Timer1
23
22
Reserved
20
MM Dply Burst
Length
7
6
Reserved
Bit
19
24
Overlay Delay Timer0
18
Reserved
17
12
MM Display FIFO
Watermark
11
Reserved
5
10
8
LM Dply Burst
Length
0
LM Dply FIFO Watermark
Description
31:28
Overlay Delay Timer1 is used to insert wait states between sets of YUVY requests to MM. The value in
this register is multiplied by 16 to determine the wait state clock count.
27:24
Overlay Delay Timer0 is used to insert wait states between any two overlay streamer requests to MM,
except between sets of YUVY. The value in this register is multiplied by 16 to determine the wait state
clock count.
23
70
Reserved
®
R
Bit
22:20
Description
MM Display Burst Length : Size in qwords of individual requests issued to memory. Use multiples of 16
QWs for tiled memory.
000 = 8 Qws
N/A if trickle feed is on.
001 = 16 Qws
010 = 24 Qws
011 = 32 Qws
100 = 40 Qws
101 = 48 Qws
110 = 56 Qws
111 = 64 QWs
19:18
Reserved
17:12
MM Display FIFO Watermark : Number of qwords stored in FIFOs, below which the DSI will generate
requests to LMI. (The value must be less than 32 and should be as recommended in the high-priority
bandwidth analysis spreadsheet.)
11
10:8
Reserved
LM Display Burst Length : Size in qwords of individual requests issued to memory. Use multiples of 16
QWs for tiled memory.
000 = 8 Qws
N/A if trickle feed is on.
001 = 16 Qws
010 = 24 Qws
011 = 32 Qws
100 = 40 Qws
101 = 48 Qws
110 = 56 Qws
111 = 64 QWs.
7:6
Reserved
5:0
LM Display FIFO Watermark : Number of qwords stored in FIFOs, below which the DSI will generate
requests to LMI. (The value must be less than 32 and should be as recommended in the high-priority
bandwidth analysis spreadsheet.)
71
®
7.6
R
I/O Control Registers
7.6.1
HVSYNC—HSYNC/VSYNC Control Register
Address offset:
Default value:
Size:
Attribute:
05000h
00000000h
32 bits
R/W
Bits 19:16 are for DPMS and DDC sync select.
DPMS MODE
HSYNC/VSYNC Control[19:16]
Power On
0000 (i.e., pulse H and V)
Standby
0010 (i.e., pulse V)
Suspend
1000 (i.e., pulse H)
Power Off
1010 (no pulse on H & V)
31
20
Reserved
19
18
17
16
VSYNC
Control
VSYNC
Data
HSYNC
Control
HSYNC
Data
15
1
Reserved
Bit
31:20
19
0
HSYNC/
VSYNC
En
Description
Reserved.
VSYNC Control. Bit 19 (VSYNC Control) and bit 18 (VSYNC Data) are used by the BIOS to
take over the sync during DDC1 communication during POST. The BIOS can force the
VSYNC data at the same time as VSYNC control enables this signal as an output, so that the
VSYNC pulse occurs on every write by the BIOS. This is done to speed up some very slow
DDC communications.
0 = Normal VSYNC output
1 = Contents of VSYNC Data will go out to VSYNC pin.
18
VSYNC Data
17
HSYNC Control
0 = Normal HSYNC output
1 = Contents of HSYNC Data will go out to HSYNC pin.
16
15:1
0
HSYNC Data.
Reserved
HSYNC/VSYNC Enable.
0 = HSync and Vsync are deactivated when the internal DAC is disabled. (Default)
1 = HSync and VSync remain active when the internal DAC is disabled via the Module PowerDown 0
Register.
72
®
R
7.7
GPIO Registers
GPIOA
General-Purpose I/O Control Register A
7.7.1
Address offset :
Default value :
Access :
Size :
05010h
00h, 00h, 000U0000b, 000U0000b
Read/Write
32 bits
This register controls the general-purpose I/O pins GPIO0 (DDCSCL pin) and GPIO1 (DDCSDA pin). These two pins
are specifically used to create a Display Data Channel (DDC) serial bus. GPIO0 = DDC Clock (DDCSCL pin) and
GPIO1 = DDC Data (DDCSDA pin). Refer to the Cspec for a description of the pin operation.
31
16
Reserved
15
13
Reserved
7
5
Reserved
12
11
10
9
8
GPIO1
data in
GPIO1
data value
GPIO1
data mask
GPIO1
direction
value
GPIO1
direction
mask
4
3
2
1
0
GPIO0
data in
GPIO0
data value
GPIO0
data mask
GPIO0
direction
value
GPIO0
direction
mask
73
®
Bit
R
Description
31:16
Reserved
15:13
Reserved
12
GPIO1 Data In (RO): This value is sampled on the GPIO1 pin as an input.
11
GPIO1 Data Value (R/W): This value should be placed on the GPIO1 pin as an output. This value only is
written into the register if GPIO1 DATA MASK also is asserted. The value will appear on the pin if this
data value actually is written to this register and the GPIO1 DIRECTION VALUE contains a value that
will configure the pin as an output.
10
GPIO1 Data Mask (R/W): This is a mask bit to determine whether the GPIO1 DATA VALUE bit should
be written into the register.
0 = Do NOT write GPIO1 Data Value bit (default).
1 = Write GPIO1 Data Value bit.
9
GPIO1 Direction Value (R/W): This value should be used to define the output enable of the GPIO1 pin.
This value only is written into the register if GPIO1 DIRECTION MASK also is asserted. The value that
will appear on the pin is defined by what is in the register for the GPIO1 DATA VALUE bit.
0 = Pin is configured as an input (default).
1 = Pin is configured as an output.
8
GPIO1 Direction Mask (R/W): This mask bit is used to determine whether the GPIO DIRECTION
VALUE bit should be written into the register.
0 = Do NOT write GPIO1 Direction Value bit (default).
1 = Write GPIO1 Direction Value bit.
7:5
Reserved
4
3
data value actually is written to this register and the GPIO0 DIRECTION VALUE contains a value that
2
GPIO0 Data Mask (R/W): This mask bit is used to determine whether the GPIO0 DATA VALUE bit
should be written into the register.
1
0
74
®
R
GPIOB
General-Purpose I/O Control Register B
7.7.2
Address offset :
Default value :
Access :
Size :
05014h
00h, 00h, 000U0000b, 000U0000b
Read/Write
32 bit
This register controls general-purpose I/O pins GPIO2 (LTVCL pin) and GPIO3 (LTVDA pin). These two pins are used
specifically to create an I2C serial bus interface. GPIO2 = I2C clock (LTVCL pin) and GPIO3 = I2C data (LTVDA pin).
Refer to the Cspec for a description of the pin operation.
31
16
Reserved
15
13
Reserved
7
5
Reserved
12
11
10
9
8
GPIO3
Data In
GPIO3
Data value
GPIO3
Data mask
GPIO3
Direction
value
GPIO3
Direction
Mask
4
3
2
1
0
GPIO2
Data In
GPIO2
Data value
GPIO2
Data mask
GPIO2
Direction
value
GPIO2
Direction
Mask
Bit
Description
31:16
Reserved
15:13
Reserved
12
11
written into the register if GPIO3 DATA MASK is also asserted. The value will appear on the pin if this
data value is actually written to this register and the GPIO3 DIRECTION VALUE contains a value that
10
GPIO3 Data Mask (R/W): This is a mask bit to determine whether the GPIO3 DATA VALUE bit should
be written into the register.
9
75
®
Bit
8
R
Description
7:5
Reserved
4
3
data value is actually written to this register and the GPIO2 DIRECTION VALUE contains a value that
2
GPIO2 Data Mask (R/W): This mask bit is used to determine whether the GPIO2 DATA VALUE bit
should be written into the register.
1
0
GPIO2 Direction Mask (R/W): This mask is used to determine whether the GPIO DIRECTION VALUE
bit should be written into the register.
76
®
R
7.8
The clock control registers are accessed by writing to the memory-mapped address offset.
The Intel 82810 Chipset has 3 PLLs for generating all the clocks. The host PLL generates the host clock, whose
frequency is controlled by an external strap. In addition, the host PLL generates the system and local memory core clock
and the graphics core clock. The hub PLL generates the clock for the hub-link unit. The display PLL generates the
display or LCD clock.
The display clock can be controlled by three blocks of registers: DCLK0, DCLK1, and DCLK2. Each display clock has
its own Display Clock i Divisor registers for M, N, and a byte of the Display & LCD Clock Divisor Select Register,
within which are P (divisor) values and which can be programmed independently. DCLK0 and DCLK1 normally are
programmed to 25.175 MHz and 28.322 MHz, respectively (VGA-compatible clocks). DCLK2 is used for non-VGA
modes.
The Display Clock i Divisor register and the appropriate byte of the Display & LCD Clock Divisor Select Register
are programmed with the loop parameters to be loaded into the clock synthesizer. The MSR[3:2] register is used to select
between DCLK0(default), DCLK1 and DCLK2. The LCD clock is selected by writing to LCD / TV Out Control [31] = 1
and [0] = 1. MSR[3:2] are ignored when this condition is TRUE.
The data written to these registers is calculated based on the reference frequency, the desired output frequency, and
characteristic VCO constraints, as described in the functional description. From the calculation, the M, N, and P values
are obtained.
7.8.1
Programming Notes
Three blocks of registers exist for programming up to three unique frequencies for the display clock. These registers are
named dclk0, dclk1 and dclk2.
Each of these blocks can be programmed independently of each other. However, only one can be selected at any point in
time in order to control the DPLL
MSR register bits 3:2 are used to determine which DCLK0,1,2 register groups will control the DPLL.
Writing to MSR register bits 3:2 also transfers the Display Clock Divisor and Display & LCD Clock Divisor Select
Register contents to the VCO register file.
Example Programming Sequence (DCLK0)
Write the Display Clock 0 Divisor register with the M-REG value and N-REG value.
Write the clock 0 byte of the Display & LCD Clock Divisor Select Register with the P-REG value.
Write the MSR register, bits 3:2 = '00', in order to select DCLK0. (NOTE: This is the default value.)
Write the MSR register, bits 3:2 = '01', in order to select DCLK1.
77
®
Write the MSR register, bit 3 = '1', to select DCLK2.
Example Programming Sequence (LCD CLK)
Write the LCD Clock Divisor register with the M-REG value and N-REG value.
Write the LCD byte of the Display & LCD Clock Divisor Select Register with the P-REG value.
Write the LCD / TV Out Control[31] = 1 and [0] = 1. MSR[3:2] are ignored when this condition is TRUE.
78
R
®
R
7.8.2
DCLK_0D—Display Clock 0 Divisor Register
Address offset:
Default value:
Attribute:
Size:
06000h–06003h
00030013h
R/W
32 bits
The Display Clock 0 Divisor register and Display & LCD Clock Divisor Select Register are programmed with the
loop parameters to be loaded into the clock synthesizer.
The data written to this register is calculated based on the reference frequency, the desired output frequency, and the
characteristic VCO constraints, as described in the datasheet.
Data is written to the Display Clock 0 Divisor register, followed by a write to clock 0 byte of the Display & LCD
Clock Divisor Select Register. The completion of the write to the clock 0 byte of the Video Clock Divisor Select
Register causes data from both registers to transfer simultaneously to the VCO register file. This prevents wild fluctuations in the VCO output during intermediate stages of a clock programming sequence.
31
26
Reserved
Bit
25
16
VCO 0 N-Divisor
15
10
Reserved
9
0
VCO 0 M-Divisor
Description
31-26
Reserved.
25-16
VCO 0 N-Divisor. N-divisor value calculated for the desired output frequency. (Default: 03h)
15-10
Reserved.
9-0
VCO 0 M-Divisor. M-divisor value calculated for the desired output frequency. (Default: 13h)
79
®
7.8.3
R
Address offset:
Default value:
Attribute:
Size:
06004h–06007h
00100053h
R/W
32 bits
Data is written to Display Clock 0 Divisor register, followed by a write to the clock 1 byte of the Display & LCD
Clock Divisor Select Register. The completion of the write to Display & LCD Clock Divisor Select Register causes
data from both registers to transfer simultaneously to the VCO register file. This prevents wild fluctuations in the VCO
output during intermediate stages of a clock programming sequence.
31
26
Reserved
Bit
16
VCO 1 N-Divisor
15
10
Reserved
9
0
VCO 1 M-Divisor
Description
31-26
Reserved
25-16
VCO 1 N-Divisor. N-Divisor value calculated for the desired output frequency. (Default: 10h)
15-10
Reserved
9-0
80
25
VCO 1 M-Divisor. M-Divisor value calculated for the desired output frequency. (Default: 53h)
®
R
7.8.4
Address offset:
Default value:
Attribute:
Size:
06008h–0600Bh
00030013h
R/W
32 bits
Data is written to Display Clock 2 Divisor register, followed by a write to the clock 2 byte of the Display & LCD
Clock Divisor Select Register. The completion of the write to the Display & LCD Clock Divisor Select Register
causes data from both registers to transfer simultaneously to the VCO register file. This prevents wild fluctuations in the
VCO output during intermediate stages of a clock programming sequence.
31
26
Reserved
Bit
25
16
VCO 2 N-Divisor
15
10
Reserved
9
0
VCO 2 M-Divisor
Description
31-26
Reserved.
25-16
VCO 2 N-Divisor. N-divisor value calculated for the desired output frequency. (Default: 03h)
15-10
Reserved.
9-0
VCO 2 M-Divisor. M-divisor value calculated for the desired output frequency. (Default: 13h)
81
®
7.8.5
R
LCD_CLKD—LCD Clock Divisor Register
Address offset:
Default value:
Attribute:
Size:
0600Ch–0600Fh
00030013h
R/W
32 bits
The LCD Clock Divisor register and Display & LCD Clock Divisor Select Register are programmed with the loop
parameters to be loaded into the clock synthesizer.
Data is written to LCD Clock Divisor register, followed by a write to the LCD clock byte of the Display & LCD Clock
Divisor Select Register. The completion of the write to Display & LCD Clock Divisor Select Register causes data
from both registers to transfer simultaneously to the VCO register file. This prevents wild fluctuations in the VCO output
during intermediate stages of a clock programming sequence.
31
26
Reserved
Bit
16
VCO LCD N-Divisor
15
10
Reserved
9
0
VCO LCD M-Divisor
Description
31:6
Reserved
25:16
VCO LCD N-Divisor. N-divisor value calculated for the desired output frequency. (Default: 03h)
15:10
Reserved
9:0
82
25
VCO LCD M-Divisor. M-divisor value calculated for the desired output frequency. (Default: 13h)
®
R
7.8.6
DCLK_0DS—Display & LCD Clock Divisor Select Register
Address offset:
Default value:
Attributes:
Size:
06010h–06013h
40404040h
R/W
32 bits
Display clock i {i=0 to 2} becomes effective after programming the appropriate byte i {i = 0 to 2}in this register. LCD
clock becomes effective after programming byte 3 in this register.
31
30
Reserved
23
Post Divisor Select LCD Clk
22
Reserved
15
20
Post Divisor Select Clk 2
14
Reserved
7
28
12
6
Reserved
4
Bit
31
30:28
27
26
25:23
27
26
Reserved
VCO
Loop Div
LCD clk
19
18
Reserved
VCO
Loop Div
clk 2
11
10
Reserved
VCO
Loop Div
clk 1
3
2
Reserved
VCO
Loop Div
clk 0
25
24
Reserved
17
16
Reserved
9
8
Reserved
1
Reserved
0
Reserved
Description
Reserved.
Post Divisor Select LCD Clock.
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
100 = Divide by 16 (default)
101 = Divide by 32
11x = Reserved
Reserved.
VCO Loop Divide LCD Clock.
0 = Divided by 4*M (default) (M = LCD Clock Divisor Register [9:0])
1 = Divided by 16*M
Reserved.
83
®
Bit
22:20
19
18
84
R
Description
Post Divisor Select clock 2.
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
Reserved.
VCO Loop Divide clock 2.
17:15
0 = Divided by 4*M (default) (M = Display Clock 2 Divisor Register [9:0])
1 = Divided by 16*M
Reserved.
14:12
11
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
Reserved.
10
9:7
1 = Divided by 16*M
Reserved.
6:4
3
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
Reserved.
2
1
1 = Divided by 16*M
Reserved.
0
Reserved. 0
®
R
7.8.7
PWR_CLKC—Power Management and Miscellaneous Clock Control
Address offset:
Default value:
Attribute:
Size:
6014h–06017h
00 00 01 03 h
R/W
32 bits
31
17
16
Reserved
15
12
11
10
9
Reserved
Reserved
.
7
2
Reserved
Bit
8
1
0
Display
Clock
PLL
VCO
Internal
DAC
Enable
Description
31:9
Reserved.
8
Reserved.
7:2
1
Display Clock PLL VCO
0 = Disable
1 = Enable (default)
0
Internal DAC Enable.
0 = Disables the internal DAC (power-down). If HSYNC/VSYNCControl[0] = 0, disables HSYNC and
VSYNC.
1 = Enables the internal DAC and does not allow disabling of HSYNC and VSYNC via HSYNC/VSYNC
Control[0]. (Default)
85
®
7.9
R
LCD / TV-Out Registers
During the LCD or TV-Out mode, the BIOS will program the following LCD / TV-out registers. These registers are 32bit, memory-mapped registers. These registers are not double-buffered and take effect when loaded. Further, this
subsystem takes into account modified CR register values during vertical blank time for centering.
This subsystem allows the timing generator to be programmed to pixel granularity. The only exception is during the
VGA pixel-doubling mode. During VGA pixel-doubling, active pixel time must be a multiple of 4 pixels to account for
centering with VGA pixel-doubling, and non-active times must be a multiple of 2 pixels clocks.
All fields are excess-0 encoded. This means that the hardware uses the value + 1, where the value is the entry in the field.
Therefore if a 0 is programmed into a field, the hardware uses the value 1 for that field.
7.9.1
HTOTAL—Horizontal Total Register
Address offset:
Default value:
Access:
Size:
31
60000h
00000000h
Read/Write
32 bits
28
Reserved
27
16
Horizontal Total Display Pixels
Bit
15
11
Reserved
10
0
Horizontal Active Display
Pixels
Description
31:28
Reserved. Read Only
27:16
Horizontal Total Display Pixels. This 12-bit field provides a horizontal total of up to 4096 pixels,
encompassing 2048 active display pixels, front/back border pixels, and the horizontal retrace period. Any
pending event (HSYNC, VSYNC) is reset at htotal.
15:11
Reserved. Read Only
10:0
Horizontal Active Display Pixels. This 11-bit field provides horizontal active display resolutions up to
2048 pixels. Note that the first horizontal active display pixel always starts at 0.
86
®
R
7.9.2
HBLANK—Horizontal Blank Register
Address offset:
Default value:
Access:
Size:
31
60004h
00000000h
Read/Write
32 bits
28
Reserved
27
16
Horizontal Blank End
Bit
15
12
Reserved
11
0
Horizontal Blank Start
Description
31:28
Reserved. Read Only
27:16
Horizontal Blank End. Horizontal blank end expressed in terms of the absolute pixel number relative to
the horizontal active display start. Note: An asserted HBlank will be deasserted when HTotal occurs,
regardless of what is programmed in the HBlank end.
15:12
Reserved. Read Only
11:0
Horizontal Blank Start. Horizontal blank start, expressed in terms of absolute pixel number relative to the
horizontal active display start
87
®
7.9.3
R
HSYNC—Horizontal Sync Register
Address offset:
Default value:
Access:
Size:
31
60008h
00000000h
Read/Write
32 bits
28
27
Reserved
16
Horizontal Sync End
15
12
11
Reserved
Bit
0
Horizontal Sync Start
Description
31:28
Reserved. Read Only
27:16
Horizontal Sync End. Horizontal sync end, expressed in terms of absolute pixel number relative to the
horizontal active display start.
Note:
1. Minimum HSYNC width is 1 pixel clock.
An asserted HSYNC will be cleared as soon as HTOTAL end is reached, regardless of the value in the
HSYNC End register.
15:12
Reserved. Read Only
11:0
Horizontal Sync Start. Horizontal sync start expressed in terms of absolute pixel number relative to the
horizontal active display start.
Note that when HSYNC Start is programmed equal to HBLANK Start, both HSYNC and HBLANK will be
asserted on the same pixel clock.
7.9.4
VTOTAL—Vertical Total Register
Address offset:
Default value:
Access:
Size:
31
6000Ch
00000000h
Read/Write
32 bits
28
Reserved
27
16
Vertical Total Display Pixels
Bit
15
11
Reserved
10
0
Vertical Active Display Pixels
Description
31:28
Reserved. Read Only
27:16
Vertical Total Display Pixels. Vertical total display lines. This 12-bit field provides a vertical total up to
4096 lines, encompassing 2048 active display lines, top/bottom border lines, and the vertical retrace
period.
15:11
Reserved. Read Only
10:0
Vertical Active Display Pixels. Vertical active display lines. This 11-bit field provides a vertical active
display resolution up to 2048 lines. Note that the first vertical active display line always starts at 0.
88
®
R
7.9.5
VBLANK—Vertical Blank Register
Address offset:
Default value:
Access:
Size:
31
60010h
00000000h
Read/Write
32 bits
28
Reserved
27
16
Vertical Blank End
Bit
15
12
Reserved
11
0
Description
31:28
Reserved. Read Only
27:16
Vertical Blank End. Vertical blank end, expressed in terms of absolute line number relative to the vertical
active display start. Note that the vertical blank will be deasserted when the vertical total occurs,
regardless of what is programmed in vertical blank end.
15:12
Reserved. Read Only
11:0
Vertical Blank Start. Vertical blank start, expressed in terms of absolute line number relative to the
vertical active display start.
89
®
7.9.6
R
VSYNC—Vertical Sync Register
Address offset:
Default value:
Access:
Size:
31
60014h
00000000h
Read/Write
32 bits
28
Reserved
27
16
Vertical Sync End
Bit
15
12
Reserved
11
0
Vertical Sync Start
Description
31:28
Reserved. Read Only
27:16
Vertical Sync End. Vertical sync end, expressed in terms of absolute line numbers relative to the vertical
active display start.
Notes:
1. Minimum VSYNC width is 2 lines. A VSYNC programmed to 1 scan line does not generate the correct
picture.
2. An asserted VSYNC will be cleared as soon as VTOTAL end is reached, regardless of the value
in the VSYNC End register.
15:12
Reserved. Read Only
11:0
Vertical Sync Start. Vertical sync start, expressed in terms of absolute line number relative to the vertical
active display start.
Notes:
1. When VSYNC start is programmed equal to VBLNK start, both VSYNC and VBLANK will be asserted
on the same pixel clock.
2. VSYNC start programmed beyond the VTOTAL end will prevent VSYNC start and VSYNC end from
occurring.
90
®
R
7.9.7
LCDTV_C—LCD/TV-Out Control Register
Address offset:
Default value:
Access:
Size:
60018h
00000000h
Read/Write
32 bits
31
30
29
28
27
LCD / TV
Out
Enable
Reserved
Centering
Enable
FP VESA
VGA
Mode
24
Reserved
23
16
(Rsvd in
GMCH)
Reserved
15
14
13
12
11
10
9
8
Reserved
FP / 740
Data
Ordering
Reserved
Reserved
VSYNC
Control
HSYNC
Control
VSYNC
Output
Control
HSYNC
Output
Control
7
6
5
4
3
2
1
0
Border
Enable
Active
Data ½
Pixel
Order
Active
Data
Polarity
VSYNC
Polarity
Control
HSYNC
Polarity
Control
BLANK#
Polarity
Control
Dot Clock
Source
Lock Dot
Clock
PLL N/M
Regs
Bit
31
Description
LCD / TV Out Enable.
1 = Enable. This bit enables the LCD / TV digital interface. The LCD / TV timing generator is jammed to
pixel 0 of the vertical front porch when this bit is 0. The timing generator may be ignored, depending
on the LCD Timing Generator Bit (29).
0 = Disable and tristate the whole interface: TVDATA[11:0], BLANK#, TVHSYNC, TVVSYNC, and
TVCLK[1:0]. CLKIN is not disabled and can be used for flat panel hot plug detection.
30
Reserved Must be programmed as 0.
29
Centering Enable.
0 = Disable. The LCD / TV timing generator controls all display timing when enabled by bit 31 above.
1 = Enable. Centers the VGA active image, as defined in the VGA CRT registers, within LCD/TV active
image.
91
®
Bit
28
R
Description
FP VESA VGA Mode
0 = Disable. Use the LCD / TV timing generator. VGA sync polarity is ignored. FP sync polarity is used.
Centering can be enabled for fixed-resolution flat panels or TVs. The Flat Panel Dot Clock PLL timing
registers must be used for both flat panels and TVs. After these registers are written, the Lock Dot
Clock PLL N/M Registers must be set to 1, which makes the dot clock PLL use only the flat panel
PLL registers.
1 = Enable. Use the VGA timing generator. VGA sync polarity is passed though, and FP sync polarity is
ignored. Centering must be disabled. Also set bit 0 of this register, Lock Dot Clock PLL N/M Regs, to
a 0, which allows normal programming of the dot clock PLL registers. This bit should be disabled
when driving a TV.
27:17
Reserved. Must be programmed as 0.
16
Reserved
15
Reserved
14
FP / 740 Data Ordering
0 = 740-compliant data ordering:
1 = Flat panel data ordering: R[7:0] ‘ G[7:4] followed by G[3:0] ‘ B[7:0].
13
Reserved
12
Reserved
11
FPVSYNC Control.
1 = FPVSYNC is disabled.
In the FP VESA VGA Mode, then this pin goes to the level of the VGA VSYNC, when disabled.
If not in the FP VESA VGA Mode, then this pin goes into the deasserted state, as specified by the VSYNC
Polarity Control field.
0 = FPVSYNC is enabled.
When in the FP VESA VGA Mode, then the VGA timing generator is the source of this signal.
When not in the FP VESA VGA Mode, then the source of this signal is this timing generator.
10
FPHSYNC Control.
1 = FPHSYNC is disabled.
If in FP VESA VGA Mode, then this pin goes to the level of the VGA HSYNC, when disabled.
If not in FP VESA VGA Mode, then this pin goes into the deasserted state, as specified by the HSYNC
Polarity Control field. 0 = FPHSYNC is enabled.
When in the FP VESA VGA Mode, then the VGA timing generator is the source of this signal
When not in the FP VESA VGA Mode, then the source of this signal is this timing generator.
9
FPVSYNC Output Control.
1 = Tristates the FPVSYNC pin.
0 = FPVSYNC is active unless LCD / TV Out Enable is deasserted.
Though this bit is provided, the GC always uses VSYNC as output.
92
®
R
Bit
8
Description
FPHSYNC Output Control.
1 = Tristates the FPHSYNC pin.
0 = FPHSYNC is active unless LCD / TV Out Enable is deasserted.
7
Border Enable.
1 = Border to the LCD / TV encoder is enabled.
0 = Border to the LCD / TV encoder is disabled.
6
Active Data Order.
1 = Reversed ½-pixel data ordering: G[3:0] ‘ B[7:0] followed by R[7:0] ‘ G[7:4].
0 = Normal ½-pixel data ordering: R[7:0] ‘ G[7:4] followed by G[3:0] ‘ B[7:0].
5
Active Data Polarity.
1 = Inverted pixel data
0 = Normal pixel data
4
VSYNC Polarity Control.
When the LCD / TV timing generator is disabled, the polarity is controlled by the VGA registers.
1 = Active HIGH
0 = Active LOW
3
HSYNC Polarity Control.
1 = Active HIGH
0 = Active LOW
2
BLANK# Polarity Control.
1 = Active HIGH
0 = Active LOW
1
Dot Clock Source.
1 = Dot clock PLL reference source is external pin = CLKIN.
0 = Dot clock PLL reference source is the default PLL source.
The CLKIN pin can be used as an Interrupt for FP hot plug detection. When the pin is used as a clock, the
interrupt signal is forced to the deassertion level.
The CLKIN / Interrupt pin is always an input. It is never disabled. An internal pull-up is active when the
pin is configured as an interrupt. When it is configured as a clock, the internal pull-up is disabled.
0
Lock Dot Clock PLL N/M Regs.
1 = Dot Clock PLL N/M registers are locked. = Use the LCD / TV PLL M/N registers and ignore the MSR
register.
0 = Dot Clock PLL N/M registers are writeable. = The MSR register controls which PLL M/N registers are
used.
When either supporting a TV encoder or a flat panel, but not in VESA VGA mode, the LCD / TV PLL M/N
registers must be set up for the proper dot clock frequency, and then this bit is written with a 1. This bit
being written with a 1 forces the dot clock PLL to look only at the LCD / TV PLL M/N registers.
93
®
7.9.8
R
OVRACT—Overlay Active Register
Address offset:
Default value:
Access:
Size:
31
6001Ch
00000000h
Read/Write
32 bits
28
Reserved
27
16
Overlay Active End
Bit
15
12
11
Reserved
0
Overlay Active Start
Description
31:27
Reserved. Read Only
26:16
Overlay Active End. This field takes into account the overlay pipeline delays for turning off the overlay at
the end of a scan line. When LCD / TV is enabled, then the overlay active end is controlled by the LCD /
TV-out timing generator, and it uses all the bits. When LCD / TV is disabled, then the overlay active end is
controlled by the VGA timing generator, and it uses bits 15:3 for character clock resolution.
15:12
Reserved. Read Only
11:0
Overlay Active Start. This field takes into account the overlay pipeline delays for lining up X = 0 to the
first active pixel. When LCD / TV is enabled, then the overlay active start is controlled by the LCD / TV-out
timing generator and all bits are used. When LCD / TV is disabled, then the overlay active start is
controlled by the VGA timing generator, and it uses bits 15:3 for the character clock.
7.9.9
BCLRPAT— Border Color Pattern Register
Address offset:
Default value:
Access:
Size:
60020h
00000000h
Read/Write
32 bits
A border is sent if border enable is on. Also, the same color will be sent during the pseudo-border period in the LCD noscalar mode.
31
0
Reserved
Bit
Green
Description
31:25
Reserved. Read Only
24:16
Red
15:8
Green
7:0
Blue
94
Red
Blue
®
R
7.10 Pixel Pipeline Control Registers
7.10.1
PIXCONF—Pixel Pipeline Configuration
Memory Offset Address:
Default:
Attributes:
70008h
00000000h
Read/Write
31
28
Reserved (0000)
23
21
Reserved (000)
15
14
13
7
27
24
Display
Gamma
Enable
Overlay
Gamma
Enable
Reserved
Reserved
19
16
Display Color Mode
11
10
9
8
Cursor
Display
Enable
Extended
Status
Read
CRT
Overscan
Color
Reserved
Palette
Addr
4
3
2
1
0
Reserved
(0)
Reserved
(System
Write 32)
Reserved
(0)
VGA
Wrap
GUI
Mode
Bit
31:28
25
12
5
Reserved (000)
26
CRT
Control
Reserved
8-Bit
DAC
Enable
20
27
Descriptions
Reserved (0000)
Display path (Graphics) Gamma Enable. (See note.)
0 = 16- and 24-bpp graphics data bypasses palette (default).
1 = 16- and 24-bpp graphics data goes through palette.
26
Overlay path Gamma Enable. (See note.)
0 = Video data bypasses palette (default).
1 = Video data goes through palette. Useful when alpha blending the overlay with the primary display, in
order to provide gamma correction for the display device. The overlay gamma correction should be set
up to un-gamma the overlay surface, bringing it into the linear space before performing the alpha
blending. Both the primary display (27 = 1) and the overlay (26=1) should be passed through the
palette after alpha blending, in order to provide proper gamma correction for the display device.
25:21
Reserved (0s)
95
®
Bit
20
R
Descriptions
CRT Control Signal Delay.
0 = CRT display enable and CRT blank are delayed for standard VGA compatibility (default).
1 = CRT display enable and CRT blank are not delayed.
This bit affects CRT Display enable and CRT Blank signal delay with respect to CRT HSYNC and CRT
VSYNC, when the standard VGA pixel pipeline is used by CRT display engine.
This bit has no effect on flat panel centering or optimized timing modes.
19:16
Display Color Mode.
0000 = CRT standard VGA text and graphics mode and 1-bit/2-bit/4-bit packed graphics mode (default).
0001 = Reserved
0010 = CRT 8-bit packed extended graphics mode
0011 = Reserved
0100 = CRT 16-bit packed (5-5-5) extended graphics mode (Targa* compatible)
0101 = CRT 16-bit packed (5-6-5) extended graphics mode (XGA compatible)
0110 = CRT 24-bit extended graphics mode compressed
0111 = CRT 24-bit extended graphics mode uncompressed. In this mode, pixels are stored only on the
lower three bytes (plane 0,1,2) of each double word, and the most-significant byte of each double
word (plane 3) is not used.
15
8-Bit DAC Enable.
0 = 6-bit DAC (default).
1 = 8-bit DAC.
14:13
12
Reserved
Hardware Cursor Display Enable.
1 = Enable.
Software should always set this bit to 1. The setting of this bit to 1 should not do any harm, even while in
the VGA mode.
11
Enable Extended Status Read Mode.
1 = Enabling this bit makes available the status of the internal state machines and the values of the red and
green data in the input holding register through the normal DAC register ports. The register ports are
redefined as follows when this bit is set:
DACMASK = Returns red input data holding value
DACWX = Returns green input data holding value
DACSTATE = Returns the status of the internal state machines in bits [7:2]
10
CRT Overscan Color
1 = Enable protected CRT overscan color (overscan[0]).
9
96
Reserved
®
R
Bit
8
Descriptions
Palette Addressing.
1 = Enable extended palette addressing (Enables access to all 8 locations.)
7:5
Reserved (000)
4
Reserved (0)
3
Reserved (0)
2
Reserved (0)
1
VGA Wrap.
0 = 256-KB wrap state (default) for memory starting at A0000
1 = Don’t wrap
0
GUI Mode.
0 = Standard VGA and extended 4-bpp, 16-color resolutions (default). Can still access memory in linear
mode.
1 = High resolution (i.e., not VGA or extended planar)
Transition from VGA modes to hi-res mode or vice-versa:
Software will turn the display engine off (screen off) using SR01[Screen Off] and will wait for at least a
couple of HSYNC periods and no more than a couple of VSYNC periods before writing to PIXCONF[0] and
turning the display on. (Since one of the isochronous streams is DRAM refresh, which is controlled by
DRAMCXH[DRAM Refresh Status], the wait should not be so long as to cause the DRAM content to
degrade.) This should ensure that all the data requested from the display engine will be out of the local
memory interface before PIXCONF is touched. In addition, while switching from hi-res to VGA or VGA to hires, the software will ensure that all of the other isochronous streams are off before programming
PIXCONF[0].
Note:
Bits [27:24] normally are not used by the graphics BIOS or by the drivers, because the gamma values are specific to a
particular display device and apply to two-color or hi-color modes (16-bit and 24-bit). It is necessary to program the
palette first with the gamma-adjusted values. There is only one palette, so if both 3:2 are set, they have the same
gamma adjustments. Typical code and typical drivers leave these bits as zero.
97
®
R
7.11 Initialization Values for VGA Registers
Mode ->
0
0*
0+
1
1*
1+
2
2*
2+
3
3*
3+
7
7+
Register
132 132
col
col
Opt 1 Opt 2
MSR
63h
A3h
67h
63h
A3h
67h
63h
A3h
67h
63h
A3h
67h
A6h
66h
6Bh
6Bh
CR00
2Dh
2Dh
2Dh
2Dh
2Dh
2Dh
5Fh
5Fh
5Fh
5Fh
5Fh
5Fh
5Fh
5Fh
A0h
9Eh
CR01
27h
27h
27h
27h
27h
27h
4Fh
4Fh
4Fh
4Fh
4Fh
4Fh
4Fh
4Fh
83h
83h
CR02
28h
28h
28h
28h
28h
28h
50h
50h
50h
50h
50h
50h
50h
50h
85h
84h
CR03
90h
90h
90h
90h
90h
90h
82h
82h
82h
82h
82h
82h
82h
82h
82h
81h
CR04
2Bh
2Bh
2Bh
2Bh
2Bh
2Bh
55h
55h
55h
55h
55h
55h
55h
55h
8Ah
8Ah
CR05
A0h
A0h
A0h
A0h
A0h
A0h
81h
81h
81h
81h
81h
81h
81h
81h
81h
9Eh
CR06
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
BFh
CR07
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
CR08
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
CR09
C7h
4Dh
4Fh
C7h
4Dh
4Fh
C7h
4Dh
4Fh
C7h
4Dh
4Fh
4Dh
4Fh
4Fh
4Fh
CR0A
06h
0Bh
0Dh
06h
0Bh
0Dh
06h
0Bh
0Dh
06h
0Bh
0Dh
0Bh
0Dh
0Dh
0Eh
CR0B
07h
0Ch
0Eh
07h
0Ch
0Eh
07h
0Ch
0Eh
07h
0Ch
0Eh
0Ch
0Eh
0Eh
0Fh
CR0C
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
CR0D
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
CR0E
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
CR0F
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
CR10
9Ch
83h
9Ch
9Ch
83h
9Ch
9Ch
83h
9Ch
9Ch
83h
9Ch
83h
9Ch
9Ch
9Ch
CR11
8Eh
85h
8Eh
8Eh
85h
8Eh
8Eh
85h
8Eh
8Eh
85h
8Eh
85h
8Eh
8Eh
8Eh
CR12
8Fh
5Dh
8Fh
8Fh
5Dh
8Fh
8Fh
5Dh
8Fh
8Fh
5Dh
8Fh
5Dh
8Fh
8Fh
8Fh
CR13
14h
14h
14h
14h
14h
14h
28h
28h
28h
28h
28h
28h
28h
28h
42h
42h
CR14
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
1Fh
0Dh
0Fh
1Fh
1Fh
CR15
96h
63h
96h
96h
63h
96h
96h
63h
96h
96h
63h
96h
63h
96h
96h
96h
CR16
B9h
BAh
B9h
B9h
BAh
B9h
B9h
BAh
B9h
B9h
BAh
B9h
BAh
B9h
B9h
B9h
CR17
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
A3h
CR18
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
SR00
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
SR01
09h
09h
08h
09h
09h
08h
01h
01h
00h
01h
01h
00h
00h
00h
01h
01h
SR02
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
SR03
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
SR04
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
03h
02h
02h
02h
GR00
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR01
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR02
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR03
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR04
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR05
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
10h
GR06
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Eh
0Ah
0Ah
0Eh
0Eh
GR07
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
GR08
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
FFh
AR00
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
98
®
R
AR01
01h
01h
01h
01h
01h
01h
01h
01h
01h
01h
01h
01h
08h
08h
01h
AR02
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
02h
08h
08h
02h
AR03
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
03h
08h
08h
03h
AR04
04h
04h
04h
04h
04h
04h
04h
04h
04h
04h
04h
04h
08h
08h
04h
AR05
05h
05h
05h
05h
05h
05h
05h
05h
05h
05h
05h
05h
08h
08h
05h
AR06
06h
14h
14h
06h
14h
14h
06h
14h
14h
06h
14h
14h
08h
08h
14h
AR07
07h
07h
07h
07h
07h
07h
07h
07h
07h
07h
07h
07h
08h
08h
07h
AR08
10h
38h
38h
10h
38h
38h
10h
38h
38h
10h
38h
38h
10h
10h
38h
AR09
11h
39h
39h
11h
39h
39h
11h
39h
39h
11h
39h
39h
18h
18h
39h
AR0A
12h
3Ah
3Ah
12h
3Ah
3Ah
12h
3Ah
3Ah
12h
3Ah
3Ah
18h
18h
3Ah
AR0B
13h
3Bh
3Bh
13h
3Bh
3Bh
13h
3Bh
3Bh
13h
3Bh
3Bh
18h
18h
3Bh
AR0C
14h
3Ch
3Ch
14h
3Ch
3Ch
14h
3Ch
3Ch
14h
3Ch
3Ch
18h
18h
3Ch
AR0D
15h
3Dh
3Dh
15h
3Dh
3Dh
15h
3Dh
3Dh
15h
3Dh
3Dh
18h
18h
3Dh
AR0E
16h
3Eh
3Eh
16h
3Eh
3Eh
16h
3Eh
3Eh
16h
3Eh
3Eh
18h
18h
3Eh
AR0F
17h
3Fh
3Fh
17h
3Fh
3Fh
17h
3Fh
3Fh
17h
3Fh
3Fh
18h
18h
3Fh
AR10
08h
08h
0Ch
08h
08h
0Ch
08h
08h
0Ch
08h
08h
0Ch
0Eh
0Eh
0Ch
0Ch
AR11
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
AR12
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
0Fh
AR13
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
00h
08h
00h
00h
00h
AR14
00h
00h
08h
00h
00h
08h
00h
00h
08h
00h
00h
08h
00h
08h
00h
99
®
R
100
R
®
8. Frame Buffer Access
The VGA frame buffer is located at A000h-BFFFh. This is the standard VGA frame buffer address.
The physical location of the frame buffer is at the top of main memory. The size can either be 512 KB or 1 MB. This is
selected in the SMRAM register, which is documented in the Initialization Registers section of this document.
The frame buffer is not stored in local memory, but it is taken from the top of main memory, as described in the
SMRAM register description.
101
®
R
102
®
R
9. VGA and Extended VGA Registers
This section describes the registers and the functional operation notations for the observable registers in the 2D section.
Each register is documented and the various bit settings defined. It is important to note that not all combinations of bit
settings result in functional operating modes. Note that these registers can be accessed via either I/O space or memory
space. The memory space addresses listed are offsets from the base memory address programmed into the MMAPA
register (PCI configuration offset 14h). For each register, the memory-mapped address offset is the same address value
as the I/O address.
9.1
General Control & Status Registers
The setup, enable, and general registers are all directly accessible by the CPU. A sub-indexing scheme is not used to read
from and write to these registers.
Name
Function
Read
I/O
ST00
ST01
VGA Input Status Register 0
VGA Input Status Register 1
Memory
Offset
3C2h
3BAh/3DAh
Write
I/O
Memory
Offset


3C2h
1
3BAh/3DAh
1

FCR
VGA Feature Control Register
3CAh
3CAh
3BAh/3DAh
MSR
VGA Miscellaneous Output
Register
3CCh
3CCh
3C2h

1
3BAh/3DAh
1
3C2h
Note:
The address selection for ST01 reads and FCR writes is dependent on CGA or MDA emulation mode, as selected via the
MSR register.
Various bits in these registers provide control over and the real-time status of the horizontal sync signal, the horizontal
retrace interval, the vertical sync signal, and the vertical retrace interval.
The horizontal retrace interval is the period during the drawing of each scan line containing active video data, when the
active video data is not being displayed. This period includes the horizontal front and back porches and the horizontal
sync pulse. The horizontal retrace interval is always longer than the horizontal sync pulse.
The vertical retrace interval is the period during which the scan lines not containing active video data are drawn. It is the
period that includes the vertical front and back porches and the vertical sync pulse. The vertical retrace interval is always
longer than the vertical sync pulse.
Display Enable is a status bit (bit 0) in VGA Input Status Register 1, that indicates when either a horizontal retrace
interval or a vertical retrace interval is taking place. In the IBM* EGA graphics system (and the ones that preceded it,
including MDA and CGA), it was important to check the status of this bit to ensure that one or the other retrace interval
was taking place, before reading from or writing to the frame buffer. In these earlier systems, reading from or writing to
the frame buffer at times other than the retrace intervals meant that the CRT controller would be denied access to the
frame buffer while the display engine was accessing the pixel data needed to draw pixels on the display. This resulted in
either “snow” or a flickering display. “Display Enable” is a poor name for this status bit, since the name suggests a
connection to the enabling or disabling of the graphics system.
103
®
9.1.1
R
ST00
Input Status 0
Default:
Attributes:
3C2h
00h
Read Only
7
5
6
CRT Int
Reserved (00)
Bit
7
4
3
RGB
Cmp /
Sen
0
Reserved (0000)
Descriptions
CRT Interrupt Pending. Note that the generation of interrupts can be enabled, through bits [4,5] of the
Vertical Retrace End Register (CR11). This ability to generate interrupts at the start of the vertical retrace
interval is a feature that is typically unused by current software. This bit is here for EGA compatibility.
0 = CRT (vertical retrace interval) interrupt is not pending.
1 = CRT (vertical retrace interval) interrupt is pending.
6:5
4
Reserved. Read as 0s.
RGB Comparator / Sense. This bit returns the state of the output of the RGB output comparator(s). The
BIOS uses this bit to determine whether the display is a color or monochrome CRT.
0 = Monochrome
1 = Color
The BIOS blanks the screen or clears the frame buffer to display only black. Next, the BIOS configures the
D-to-A converters and the comparators to test for the presence of a color display. Finally, if the BIOS does
not detect a color display, it tests for the presence of a monochrome display. The result of each such test
is read via this bit.
3:0
104
®
R
9.1.2
ST01
Input Status 1
Default:
Attributes:
3BAh/3DAh
00h
Read Only
The address selection is dependent on the CGA or MDA emulation mode, as selected via the MSR register.
7
6
Reserved
(0)
Reserved
(0)
Bit
5
4
Video Feedback
Vertical
Retrace
Reserved (as per VGA specification). Read as 0s.
6
Reserved. Read as 0.
3
2
1
Reserved (00)
0
Display
Enable
Descriptions
7
5:4
3
Video Feedback 1, 0. These diagnostic video bits are selected by means of the Color Plane Enable
Register. These bits are programmably connected to 2 of the 8 color bits sent to the palette. Bits 4 and 5 of
the Color Plane Enable Register (AR12) selects which two of the 8 possible color bits become connected to
these 2 bits of this register. The current software normally does not use these 2 bits. They exist for EGA
compatibility.
Vertical Retrace/Video.
0 = VSYNC inactive (Indicates that a vertical retrace interval is not taking place.)
1 = VSYNC active (Indicates that a vertical retrace interval is taking place.)
Note:
Bits 4 and 5 of the Vertical Retrace End Register (CR11) can program this bit to generate an interrupt at
the start of the vertical retrace interval. This ability to generate interrupts at the start of the vertical
retrace interval is a feature that is largely unused by current software.
2:1
0
Display Enable Output.
0 = DE inactive. Active display area data is being drawn on the display. Neither a horizontal retrace interval
nor a vertical retrace interval is currently taking place.
1 = DE active. Either a horizontal retrace interval or a vertical retrace interval is currently taking place.
105
®
9.1.3
R
FCR
Feature Control
Default:
Attributes:
3BAh/3DAh Write; 3CAh Read
00h
See address above
The address selection for reads is dependent on the CGA or MDA emulation mode, as selected via the MSR register.
7
4
Reserved (0000)
Bit
7:4
3
3
VSYNC
Control
2
0
Reserved (000)
Descriptions
VSYNC Control.
0 = Vsync output on the VSYNC pin (default).
1 = Logical OR of VSync and Display Enable output on the VSYNC pin. This capability is not typically very
useful, but is provided for IBM compatibility.
2:0
106
®
R
9.1.4
MSR
Miscellaneous Output
Default:
Attributes:
3C2h  Write; 3CCh Read
00h
See Address above
7
6
5
4
VSYNC
Polarity
HSYNC
Polarity
Page
Select
Reserved
(0)
Bit
7
3
2
Clock Select
1
0
A0000−B
FFFFh
Acc En
I/O
Address
Descriptions
CRT VSync Polarity.
0 = Positive Polarity (default)
1 = Negative Polarity
6
CRT HSync Polarity.
0 = Positive Polarity (default)
1 = Negative Polarity
5
Page Select. In Odd/Even Memory Map Mode 1 (GR6), this bit selects the upper or lower 64-KB page in display memory for CPU access:
0 = Upper page (default)
1 = Lower page
Selects between two 64-KB pages of frame buffer memory during standard VGA odd/even modes (modes
0h through 5h). Bit 1 of register GR06 can also program this bit in other modes. Note that this bit is always
set to 1 by the driver software.
4
3:2
Clock Select. These bits usually select the dot clock source for the CRT interface. The bits select the dot
clock in standard VGA modes.
00 = CLK0, 25.175 MHz (for standard VGA modes with 640-pixel horizontal resolution) (default)
01 = CLK1, 28.322 MHz (for standard VGA modes with 720-pixel horizontal resolution)
1x = CLK2 (Left “reserved” in standard VGA, used for all extended modes 6 MHz - 135 MHz.)
1
A0000−
−BFFFFh Access Enable. VGA Compatibility bit enables access to local video memory (frame buffer)
at A0000−BFFFFh. When it is disabled, accesses to system memory are blocked in this region (by not
asserting DEVSEL#). This bit does not block CPU access to the video linear frame buffer at other addresses.
0 = Prevent CPU access to frame buffer (default).
1 = Allow CPU access to frame buffer.
0
I/O Address Select. This bit selects 3Bxh or 3Dxh as the I/O address for the CRT controller registers, the
Feature Control Register (FCR), and Input Status Register 1 (ST01). Presently ignored (whole range is
claimed), but will “ignore” 3Bx for color configuration or 3Dx for monochrome.
0 = Select 3Bxh I/O address (MDA emulation) (default).
1 = Select 3Dxh I/O address (CGA emulation).
107
®
R
Note:
In standard VGA modes, bits 7 and 6 indicate which of the three standard VGA vertical resolutions the standard VGA
display should use. All extended modes, including those with a vertical resolution of 480 scan lines, use a setting of 0 for
both of these bits. This setting was “reserved” in the VGA standard.
Table 4. CRT Display Sync Polarities
V
H
Display
Horizontal Frequency
Vertical Frequency
P
P
>480 Line
Variable
Variable
P
P
200 Line
15.7 KHz
60 Hz
N
P
350 Line
21.8 KHz
60 Hz
P
N
400 Line
31.5 KHz
70 Hz
N
N
480 Line
31.5 KHz
60 Hz
9.2
Sequencer Registers
The sequencer registers are accessed via either I/O space or memory space. To access the registers, the VGA Sequencer
Index Register (SRX) at I/O address 3C4h (or memory address 3C4h) is written with the index of the desired register,
and then the desired register is accessed through the data port for the sequencer registers at I/O address 3C5 (or memory
address 3C5).
9.2.1
SRX
Sequencer Index
Default:
Attributes:
3C4h
00h
Read/Write
7
3
Reserved (00000)
Bit
2
0
Sequencer Index
Descriptions
7:3
2:0
Sequencer Index. This field contains the 3-bit sequencer index value used to access sequencer data registers at indices 0 through 7.
Notes:
1. SR02 is referred to in the VGA standard as the Map Mask Register. However, “map” is used with
multiple meanings in the VGA standard and was, therefore, deemed too confusing. Hence, the reason
for calling it the Plane Mask Register.
2. SR07 is a standard VGA register that was not documented by IBM. It is not an graphics controller
extension.
108
®
R
9.2.2
SR00
Sequencer Reset
Default:
Attributes:
3C5h(Index=00h)
00h
Read/Write
7
2
Reserved (000000)
Bit
7:2
0
Reserved. Scratch bit required for VGA compatibility
Reserved
(scratch
bit)
SR01
Clocking Mode
Default:
Attributes:
7
3C5h (Index=01h)
00h
Read/Write
6
Reserved (00)
5
4
3
2
1
0
Screen
Off
Shift 4
Dot
Clock
Divide
Shift
Load
Reserved
(0)
8/9 Dot
Clocks
Bit
5
Reserved
(scratch
bit)
Reserved.
Reserved. Scratch bit required for VGA compatibility
7:6
0
Descriptions
1
9.2.3
1
Descriptions
Screen Off.
0 = Normal operation (default)
1 = Disables video output (i.e., blanks the screen) and turns off the picture-generating logic. This allows the
full memory bandwidth to be available for CPU accesses. Synchronization pulses to the display,
however, are maintained. Setting this bit to 1 can be used as a way to more rapidly update the frame
buffer.
4
Shift 4.
0 = Load video shift registers every 1 or 2 character clocks (depending on bit 2 of this register) (default).
1 = Load shift registers every 4th character clock.
109
®
R
Bit
Descriptions
3
Dot Clock Divide. Setting this bit to 1 divides the dot clock by two and stretches all timing periods. This bit
is used in standard VGA 40-column text modes to stretch timings in order to create horizontal resolutions of
either 320 or 360 pixels (as opposed to 640 or 720 pixels, normally used in standard VGA 80-column text
modes).
0 = Sequencer master clock output on the PCLK pin (used for 640 (720) pixel modes). Pixel clock is left
unaltered (default).
1 = Pixel clock divided by 2 output on the PCLK pin (used for 320 (360) pixel modes)
2
Shift Load. Bit 4 of this register must be 0 for this bit to be effective.
0 = Load video data shift registers every character clock (default).
1 = Load video data shift registers every other character clock.
1
0
8/9 Dot Clocks. This bit determines whether a character clock is 8 or 9 dot clocks long.
0 = 9 dot clocks (9 horizontal pixels) per character in text modes with a horizontal resolution of 720 pixels
(default)
1 = 8 dot clocks (8 horizontal pixels) per character in text modes with a horizontal resolution of 640 pixels
9.2.4
SR02
Plane/Map Mask
Default:
Attributes:
7
4
Reserved
Bit
7:4
3:0
3C5h (Index=02h)
00h
Read/Write
3
0
Memory Planes Processor Write Access Enable
Descriptions
Memory Planes [3:0] Processor Write Access Enable. In both the Odd/Even Mode and the Chain 4
Mode, these bits still control access to the corresponding color plane.
0 = Disable.
1 = Enable.
Note:
This register is referred to in the VGA standard as the Map Mask Register. However, “map” is used
with multiple meanings in the VGA standard and was, therefore, considered too confusing. Hence, the
reason for calling it the Plane Mask Register.
110
®
R
9.2.5
SR03
Character Font
Default:
Attributes:
7
3C5h (index=03h)
00h
Read/Write
6
Reserved (00)
5
4
Char
Map A
Select
(bit 0)
Char
Map B
Select
(bit 0)
Bit
7:6
3:2,5
2
Character Map A
Select (bits 2 and 1)
1
0
Character Map B
Select (bits 2 and 1)
Descriptions
Character Map Select Bits for Character Map B. These three bits are used to select the character map
(character generator tables) to be used as the secondary character set (font). Note that the numbering of
the maps is not sequential.
Bit [3:2, 5]
00,0
00,1
01,0
01,1
10,0
10,1
11,0
11,1
1:0,4
3
Map Number
0
4
1
5
2
6
3
7
Table Location
1st 8 KB of plane 2 at offset 0 (default)
2nd 8 KB of plane 2 at offset 8 K
3rd 8 KB of plane 2 at offset 16 K
4th 8 KB of plane 2 at offset 24 K
Character Map Select Bits for Character Map A. These three bits are used to select the character map
(character generator tables) to be used as the primary character set (font). Note that the numbering of the
maps is not sequential.
Bit [1:0,4]
0,00
0,01
0,10
0,11
1,00
1,01
1,10
1,11
Map Number
0
4
1
5
2
6
3
7
Table Location
1st 8 KB of plane 2 at offset 0 (default)
2nd 8 KB of plane 2 at offset 8 K
3rd 8 KB of plane 2 at offset 16 K
NOTES:
In text modes, bit 3 of the video data’s attribute byte normally controls the foreground intensity. This bit may be
redefined to control switching between character sets. This latter function is enabled whenever there is a difference in the
values of the Character Font Select A and the Character Font Select B bits. If the two values are the same, the character
select function is disabled and attribute bit 3 controls the foreground intensity.
Bit 1 of the Memory Mode Register (SR04) must be set to 1 for the character font select function of this register to be
active. Otherwise, only character maps 0 and 4 are available.
111
®
R
SR04
Memory Mode Register
9.2.6
Default:
Attributes:
3C5h (index=04h)
00h
Read/Write
7
4
Reserved (0000)
Bit
7:4
3
3
2
1
0
Chain 4
Odd/Even
Extended
Memory
Reserved
(0)
Description
Chain 4 Mode. The selections made by this bit affect both CPU read and write accesses to the frame
buffer.
0 = The manner in which the frame buffer memory is mapped is determined by the setting of bit 2 of this
register (default).
1 = The frame buffer memory is mapped in such a way that the function of address bits 0 and 1 are altered
so that they select planes 0 through 3.
2
Odd/Even Mode. Bit 3 of this register must be set to 0 for this bit to be effective. The selections made by
this bit affect only CPU writes to the frame buffer.
0 = The frame buffer memory is mapped so that the function of address bit 0 is such that even addresses
select planes 0 and 2 and odd addresses select planes 1 and 3 (default).
1 = Addresses sequentially access data within a bitmap, and the choice of which map is accessed is made
according to the value of the Plane Mask Register (SR02).
1
Extended Memory Enable. This bit must be set to 1 to enable the selection and use of character maps in
plane 2 via the Character Map Select Register (SR03).
0 = Disable CPU accesses to more than the first 64 KB of VGA standard memory (default).
1 = Enable CPU accesses to the rest of the 256 KB total VGA memory beyond the first 64 KB.
0
112
®
R
SR07
Horizontal Character Counter Reset
9.2.7
Default:
Attributes:
3C5h (index=07h)
00h
Read/Write
Writing this register with any data causes the horizontal character counter to be held in reset (i.e., the character counter
output will remain 0), until a write occurs to any other sequencer register location with SRX set to an index of 0 through
6.
The vertical line counter is clocked by a signal derived from the horizontal display enable (which does not occur if the
horizontal counter is held in reset). Therefore, if a write occurs to this register during the vertical retrace interval, both
the horizontal and vertical counters will be set to 0. A write to any other sequencer register location (with SRX set to an
index of 0 through 6) may then be used to start both counters with reasonable synchronization to an external event, via
software control.
This is a standard VGA register that was not documented by IBM.
Bit
Description
Horizontal Character Counter.
7:0
9.3
Graphics Controller Registers
The graphics controller registers are accessed via either I/O space or memory space. To access the registers, the VGA
Graphics Controller Index Register at I/O address 3CEh (or memory address 3CEh) is written with the index of the
desired register, and then the desired register is accessed through the data port for the graphics controller registers at I/O
address 3CFh (or memory address 3CFh).
GRX
GRX Graphics Controller Index Register
9.3.1
Default:
Attributes:
3CEh
000UUUUb (U = undefined)
Read/Write
7
4
3
Reserved (0000)
Bit
0
Graphics Controller Register Index
Description
7:4
3:0
Sequencer Register Index. This field selects any one of the graphics controller registers (GR[00:0F]) to be
accessed via the data port at I/O location 3CFh.
113
®
R
GR00
Set/Reset Register
9.3.2
Default:
Attributes:
3CFh (index=00h)
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
1
0
Set/Reset
Plane 3
Set/Reset
Plane 2
Set/Reset
Plane 1
Set/Reset
Plane 0
Description
7:4
3:0
Set/Reset Plane [3:0]. When the Write Mode bits (bits 0 and 1) of the Graphics Mode Register (GR05) are
set to select Write Mode 0, all 8 bits of each byte of each memory plane are set to either 1 or 0, as
specified in the corresponding bit in this register, if the corresponding bit in the Enable Set/Reset Register
(GR01) is set to 1.
When the Write Mode bits (bits 0 and 1) of the Graphics Mode Register (GR05) are set to select Write
Mode 3, all CPU data written to the frame buffer is rotated, logically ANDed with the contents of the Bit
Mask Register (GR08), and then treated as the addressed data’s bit mask, while the value of these four
bits of this register are treated as the color value.
GR01
Enable Set/Reset Register
9.3.3
Default:
Attributes:
3CFh (Index=01h)
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
1
0
Enable
Set/ Reset
Pln 3
Enable
Set/ Reset
Pln 2
Enable
Set/ Reset
Pln 1
Enable
Set/ Reset
Pln 0
Description
7:4
3:0
Enable Set/Reset Plane [3:0]. This register works in conjunction with the Set/Reset Register (GR00). The
Write Mode bits (bits 0 and 1) must be set for Write Mode 0 for this register to have any effect.
0 = The corresponding memory plane can be read from or written to by the CPU, without any special
bitwise operations taking place.
1 = The corresponding memory plane is set to 0 or 1, as specified in the Set/Reset Register (GR00).
114
®
R
GR02
Color Compare Register
9.3.4
Default:
Attributes:
3CFh (Index=02h)
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
1
0
Color
Compare
Plane 3
Color
Compare
Plane 2
Color
Compare
Plane 1
Color
Compare
Plane 0
Description
7:4
3:0
Color Compare Plane [3:0]. When the Read Mode bit (bit 3) of the Graphics Mode Register (GR05) is set
to select Read Mode 1, all 8 bits (of each byte of each of the 4 memory planes of the frame buffer
corresponding to the address from which a CPU read access is being performed) are compared with the
corresponding bits in this register (if the corresponding bit in the Color Don’t Care Register (GR07) is set to
1). The 8-bit value that the CPU receives from the read access shows the result of this comparison,
wherein value of 1 in a given bit position indicates that all of the corresponding bits in the bytes across all of
the memory planes that were included in the comparison had the same value as their memory plane’s
respective bits in this register.
GR03
Data Rotate Register
9.3.5
Default:
Attributes:
7
3CFh (Index=03h)
0Uh (U = undefined)
Read/Write
5
Reserved
Bit
4
3
Function Select
2
0
Rotate Count
Description
7:5
4:3
Function Select. These bits specify the logical function (if any) to be performed on data that is meant to be
written to the frame buffer (using the contents of the memory read latch), just before it is actually stored in
the frame buffer at the intended address location.
00 = Data being written to the frame buffer remains unchanged and is simply stored in the frame buffer.
01 = Data being written to the frame buffer is logically ANDed with the data in the memory read latch,
before it is actually stored in the frame buffer.
10 = Data being written to the frame buffer is logically ORed with the data in the memory read latch, before
it is actually stored in the frame buffer.
11 = Data being written to the frame buffer is logically XORed with the data in the memory read latch,
before it is actually stored in the frame buffer.
2:0
Rotate Count. These bits specify the number of bits to the right to rotate any data that is meant to be
written to the frame buffer, just before it is actually stored in the frame buffer at the intended address
location.
115
®
R
GR04
Read Plane Select Register
9.3.6
Default:
Attributes:
3CFh (Index=04h)
0Uh (U = undefined)
Read/Write
7
2
1
Reserved (000000)
Bit
0
Read Plane Select
Description
7:2
1:0
Read Plane Select. These two bits select the memory plane from which the CPU reads data in Read Mode
0. In Odd/Even Mode, bit 0 of this register is ignored. In Chain 4 Mode, both bits 1 and 0 of this register are
ignored. The four memory planes are selected as follows:
00 = Plane 0
01 = Plane 1
10 = Plane 2
11 = Plane 3
These two bits also select which of the four memory read latches may be read via the Memory Read Latch
Data Register (CR22). The choice of memory read latch corresponds to the choice of plane specified in the
table above. The Memory Read Latch Data Register and this additional function served by 2 bits are
features of the VGA standard that were never documented by IBM.
9.3.7
GR05
Graphics Mode Register
Default:
Attributes:
7
Reserved
(0)
6
Shift Register Control
Bit
7
116
3CFh (Index=05h)
0UUU U0UUb (U = undefined)
Read/Write
5
4
3
2
Odd/Even
Read
Mode
Reserved
(0)
Description
1
0
Write Mode
®
R
Bit
Description
6:5
Shift Register Control. In standard VGA modes, pixel data is transferred from the 4 graphics memory
planes to the palette via a set of 4 serial output bits. These 2 bits of this register control the format in which
data in the 4 memory planes is serialized for these transfers to the palette.
Bits [6:5]=00
One bit of data at a time from parallel bytes in each of the 4 memory planes is transferred to the palette via
the 4 serial output bits, with 1 of each serial output bit corresponding to a memory plane. This provides a 4bit value on each transfer for 1 pixel, making possible a choice of 1 of 16 colors per pixel.
Serial
Out
1st Xfer
2nd Xfer
3rd Xfer
4th Xfer
5th Xfer
6th Xfer
7th Xfer
8th Xfer
Bit 3
plane 3 bit 7
plane 3 bit 6
plane 3 bit 5
plane 3 bit 4
plane 3 bit 3
plane 3 bit 2
plane 3 bit 1
plane 3 bit 0
Bit 2
plane 2 bit 7
plane 2 bit 6
plane 2 bit 5
plane 2 bit 4
plane 2 bit 3
plane 2 bit 2
plane 2 bit 1
plane 2 bit 0
Bit 1
plane 1 bit 7
plane 1 bit 6
plane 1 bit 5
plane 1 bit 4
plane 1 bit 3
plane 1 bit 2
plane 1 bit 1
plane 1 bit 0
Bit 0
plane 0 bit 7
plane 0 bit 6
plane 0 bit 5
plane 0 bit 4
plane 0 bit 3
plane 0 bit 2
plane 0 bit 1
plane 0 bit 0
Bits [6:5]=01
Two bits of data at a time from parallel bytes in each of the 4 memory planes are transferred to the palette,
in a pattern that alternates per byte between memory planes 0 and 2 and memory planes 1 and 3. First the
even-numbered and odd-numbered bits of a byte in memory plane 0 are transferred via serial output bits 0
and 1, respectively, while the even-numbered and odd-numbered bits of a byte in memory plane 2 are
transferred via serial output bits 2 and 3. Next, the even-numbered and odd-numbered bits of a byte in
memory plane 1 are transferred via serial output bits 0 and 1, respectively, while the even-numbered and
odd-numbered bits of memory plane 3 are transferred via serial out bits 1 and 3. This provides a pair of 2bit values (one 2-bit value for each of 2 pixels) on each transfer, making possible a choice of 1 of 4 colors
per pixel.
Serial
Out
1st Xfer
2nd Xfer
3rd Xfer
4th Xfer
5th Xfer
6th Xfer
7th Xfer
8th Xfer
Bit 3
plane 2 bit 7
plane 2 bit 5
plane 2 bit 3
plane 2 bit 1
plane 3 bit 7
plane 3 bit 5
plane 3 bit 3
plane 3 bit 1
Bit 2
plane 2 bit 6
plane 2 bit 4
plane 2 bit 2
plane 2 bit 0
plane 3 bit 6
plane 3 bit 4
plane 3 bit 2
plane 3 bit 0
Bit 1
plane 0 bit 7
plane 0 bit 5
plane 0 bit 3
plane 0 bit 1
plane 1 bit 7
plane 1 bit 5
plane 1 bit 3
plane 1 bit 1
Bit 0
plane 0 bit 6
plane 0 bit 4
plane 0 bit 2
plane 0 bit 0
plane 1 bit 6
plane 1 bit 4
plane 1 bit 2
plane 1 bit 0
This alternating pattern is meant to accommodate the use of the odd/even mode of organizing the 4
memory planes, which is used by standard VGA modes 2h and 3h.
Bits [6:5]=1x
Four bits of data at a time from parallel bytes in each of the 4 memory planes are transferred to the palette
in a pattern that iterates per byte through memory planes 0 through 3. First the 4 most-significant bits of a
byte in memory plane 0 are transferred via the 4 serial output bits, followed by the 4 least-significant bits of
the same byte. Next, the same transfers occur from the parallel byte in memory planes 1, 2, and lastly 3.
Each transfer provides either the upper or lower half of an 8-bit value for the color for each pixel, making
possible a choice of 1 of 256 colors per pixel.
Serial
Out
1st Xfer
2nd Xfer
3rd Xfer
4th Xfer
5th Xfer
6th Xfer
7th Xfer
8th Xfer
Bit 3
plane 0 bit 7
plane 0 bit 3
plane 1 bit 7
plane 1 bit 3
plane 2 bit 7
plane 2 bit 3
plane 3 bit 7
plane 3 bit 3
Bit 2
plane 0 bit 6
plane 0 bit 2
plane 1 bit 6
plane 1 bit 2
plane 2 bit 6
plane 2 bit 2
plane 3 bit 6
plane 3 bit 2
Bit 1
plane 0 bit 5
plane 0 bit 1
plane 1 bit 5
plane 1 bit 1
plane 2 bit 5
plane 2 bit 1
plane 3 bit 5
plane 3 bit 1
Bit 0
plane 0 bit 4
plane 0 bit 0
plane 1 bit 4
plane 1 bit 0
plane 2 bit 4
plane 2 bit 0
plane 3 bit 4
plane 3 bit 0
This pattern is meant to accommodate mode 13h, a standard VGA 256-color graphics mode.
117
®
Bit
4
R
Description
Odd/Even Mode.
0 = Addresses sequentially access data within a bitmap, and the choice of which map is accessed is made
according to the value of the Plane Mask Register (SR02).
1 = The frame buffer is mapped so that the function of address bit 0 is such that even addresses select
memory planes 0 and 2 and odd addresses select memory planes 1 and 3.
Note:
This works in a way that is the inverse of (and is normally set to be the opposite of) bit 2 of the Memory
Mode Register (SR02).
3
Read Mode.
0 = During a CPU read from the frame buffer, the value returned to the CPU is data from the memory
plane selected by bits 1 and 0 of the Read Plane Select Register (GR04).
1 = During a CPU read from the frame buffer, all 8 bits (of the byte in each of the 4 memory planes
corresponding to the address from which a CPU read access is being performed) are compared to the
corresponding bits in this register (if the corresponding bit in the Color Don’t Care Register (GR07) is
set to 1). The 8-bit value that the CPU receives from the read access shows the result of this
comparison. A value of 1 in a given bit position indicates that all of the corresponding bits in the bytes
across all 4 of the memory planes that were included in the comparison had the same value as their
memory plane’s respective bits in this register.
2
1:0
Write Mode.
00 = Write Mode 0  During a CPU write to the frame buffer, the addressed byte in each of the 4 memory
planes is written with the CPU write data, after it has been rotated by the number of counts specified
in the Data Rotate Register (GR03). If, however, the bit(s) in the Enable Set/Reset Register (GR01)
corresponding to one or more of the memory planes is set to 1, then the data stored in the
corresponding bits in the Set/Reset Register (GR00) will be written to those memory planes.
01 = Write Mode 1  During a CPU write to the frame buffer, the data stored in the memory read latches
is written to the addressed byte in each of the 4 memory planes. (The memory read latches store an
unaltered copy of the data last read from any location in the frame buffer.)
10 = Write Mode 2  During a CPU write to the frame buffer, the least-significant 4 data bits of the CPU
write data are treated as the color value for the pixels in the addressed byte in all 4 memory planes.
The 8 bits of the Bit Mask Register (GR08) are used to selectively enable or disable the ability to write
to the corresponding bit in each of the 4 memory planes that correspond to a given pixel. A setting of
0 in a bit in the Bit Mask Register at a given bit position causes the bits in the corresponding bit
positions in the addressed byte in all 4 memory planes to be written with value of their counterparts in
the memory read latches. A setting of 1 in a Bit Mask Register at a given bit position causes the bits
in the corresponding bit positions in the addressed byte in all 4 memory planes to be written with the
4 bits taken from the CPU write data, thereby causing the pixel corresponding to these bits to be set
to the color value.
11 = Write Mode 3  During a CPU write to the frame buffer, the CPU write data is logically ANDed with
the contents of the Bit Mask Register (GR08). The result of this ANDing is treated as the bit mask
used in writing the contents of the Set/Reset Register (GR00), which are written to addressed byte in
all 4 memory planes.
118
®
R
GR06
Miscellaneous Register
9.3.8
Default:
Attributes:
3CFh (Index=06h)
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
Memory Map Mode
1
0
Chain
Odd/Even
Graphics /
Text Mode
Description
7:4
3:2
Memory Map Mode. These 2 bits control the mapping of the frame buffer into the CPU address space, as
follows:
Bit [3:2] Frame Buffer Address Range
A0000h − BFFFFh
A0000h − AFFFFh
B0000h − B7FFFh
B8000h − BFFFFh
00
01
10
11
NOTE
This function is in both standard VGA modes and extended modes that do not provide linear frame
buffer access.
Software must set to the proper value.
1
Chain Odd/Even. This bit provides the ability to alter the interpretation of address bit A0, so that it may be
used in selecting between the odd-numbered memory planes (planes 1 and 3) and the even-numbered
memory planes (planes 0 and 2).
0 = A0 functions normally.
1 = A0 is switched with a high order address bit, in terms of how it is used in address decoding. The result
is that A0 is used to determine which memory plane is being accessed
(A0=0 for planes 0 and 2, A0=1 for planes 1 and 3).
0
Graphics/Text Mode.
0 = Text mode
1 = Graphics mode
119
®
R
GR07
Color Don’t Care Register
9.3.9
Default:
Attributes:
3CFh (Index=07h)
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
1
0
Ignore
Color
Plane 3
Ignore
Color
Plane 2
Ignore
Color
Plane 1
Ignore
Color
Plane 0
Description
7:4
3:0
Ignore Color Plane [3:0]. Note that these bits have effect only when bit 3 of the Graphics Mode Register
(GR05) is set to 1 to select Read Mode 1.
0 = The corresponding bit in the Color Compare Register (GR02) will not be included in color comparisons.
1 = The corresponding bit in the Color Compare Register (GR02) is used in color comparisons.
9.3.10
GR08
Bit Mask Register
Default:
Attributes:
Bit
7:0
3CFh (Index=08h)
Undefined
Read/Write
Description
Bit Mask.
0 = The corresponding bit in each of the 4 memory planes is written with the corresponding bit in the
memory read latches.
1 = The manipulation of the corresponding bit in each of the 4 memory planes via other mechanisms is
enabled.
NOTES
This bit mask applies simultaneously to any writes to the addressed byte of any or all of the 4 memory
planes.
This bit mask is applicable to any data written into the frame buffer by the CPU, including data that is also
subject to rotation, logical functions (AND, OR, XOR), and set/reset. To perform a proper read-modifywrite cycle into the frame buffer, each byte first must be read from the frame buffer by the CPU, which
will cause it to be stored in the memory read latches, after which this Bit Mask Register must be set
and the new data then are written into the frame buffer by the CPU.
120
®
R
9.3.11
GR10
Address Mapping
Default:
Attributes:
7
3CFh (Index=10h)
00h
R/W
5
Reserved (0000)
Bit
7:5
4
4
3
2
1
0
Paging to
LM
VGA
Buffer
/Memory
Map
Packed
Mode
Enbl
Linear
Mapping
Page
Mapping
Description
Reserved (000)
Page to Local Memory Enable:
Used Only if GR10(0) = 1 {paging enabled} and (GR10(1) = 1 or GR10(2) = 1) {either Packed mode
or Linear mode is enabled) and GR10(3) = 0 {VGA buffer selected}.
0 = Page to VGA buffer.
1 = Page to physical local memory.
3
VGA Buffer/Memory Map Select.
0 = VGA buffer (default)
1 = Memory map
2
Packed Mode Enable.
0 = Address and data translation are bused register settings (default).
1 = Forced extended pack pixel address translation.
In page mapping mode, register GR06 selects the video memory address.
1
Linear Mapping (PCI).
1 = Enable
0
Page Mapping Enable. This mode allows the mapping of the VGA space allocated in main memory (nonlocal video memory) mode or all of local memory space through the [A0000:AFFFF] window (using bit 4 of
this register), which is a 64-KB page. An internal address is generated using GR11[6:0] as the address
line [22:16] extension to A[15:2].
1 = Enable
121
®
R
Table 5. VGA Address Range
GR10[
2]
GR10
[1]
GR10
[0]
Note 1
Address Range (see note 2)
A0000-AFFFF Range (No
GTT)
B0, B8 Ranges (No GTT)
0
0
0
Std VGA
xlations
0
0
1
Paging and
VGA xlation
VGA controller, paged by GR11
0
1
0
No paging, no
VGA xlations
0
1
1
Paging, no
VGA xlations
1
0
0
No paging, no
VGA xlations
1
0
1
Paging, no
VGA xlations
1
1
0
No paging, no
VGA xlations
1
1
1
Paging, no
VGA xlations
Bypass VGA paged by GR11
NOTES:
GR10[2:0]: 001 should not be used for paging, as all the VGA registers need to be setup correctly. An access thru
A0000 range is paged by GR11. Note: prefetch refers to cache line size access to GM vs. without prefetch (i.e.,
QW).
VGA address range, selected by GR06; graphics range selected through graphics base address register in
configuration space. Access to VGA range does not require a translation table and VGA range paging allows
access to all of local memory, if it is set up with bit 4 of this register or to all the VGA main memory space.
Access to graphics range requires GTT to be set up and will result in a prefetch, unless prefetch is disabled.
Access to VGA range will not result in prefetch.
BIOS should access local memory through the "back door" mechanism, by setting gr10 = 17h, gr11 = 0, and gr6 = 0,
only when local memory has been enabled (MMADR+3000h). Otherwise, the system will hang forever in a
snoop stall.
122
®
R
9.3.12
GR11
Page Selector
Default :
Attributes:
3CFh (Index=11h)
00h
R/W
Bit
7:0
9.3.13
Description
Page Select. Selects a 64-KB window within VGA space in NLVM mode or all of local memory when
page mapping is enabled (GR10[0]=1). In addition, this register is used for page selection of memorymapped register addresses.
GR[14:1F]
Software Flags x
Default:
Attribute:
Bit
7:0
3CFh (Index = 14h-1fh)
00
R/W
Description
Software Flags. Used as scratch pad space in BIOS. These have no effect on H/W.
123
®
9.4
R
Attribute Controller Registers
Unlike the other sets of indexed registers, the attribute controller registers are not accessed through a scheme employing
an entirely separate index and data ports. I/O address 3C0h (or memory address 3C0h) is used both as the read and write
for the index register and as the write address for the data port. I/O address 3C1h (or memory address 3C1h) is the read
address for the data port.
To write to one of the attribute controller registers, the index of the desired register must be written to I/O address 3C0h
(or memory address 3C0h), and then the data is written to the same I/O (memory) address. A flip-flop alternates with
each write to I/O address 3C0h (or memory address 3C0h), in order to toggle its function between writing the index to
writing the actual data. This flip-flop may be deliberately set so that I/O address 3C0h (or memory address 3C0h) is set
to write to the index (which provides a way to set it to a known state), by performing a read operation from Input Status
Register 1 (ST01) at I/O address 3BAh (or memory address 3BAh) or 3DAh (or memory address 3DAh), depending on
whether the graphics system has been set to emulate an MDA or a CGA, as per MSR[0].
To read from one of the attribute controller registers, the index of the desired register must be written to I/O address
3C0h (or memory address 3C0h), and then the data is read from I/O address 3C1h (or memory address 3C1h). A read
operation from I/O address 3C1h (or memory address 3C1h) does not reset the flip-flop to writing to the index. Only a
write to 3C0h (or memory address 3C0h) or a read from 3BAh or 3DAh (or memory address 3BAh or 3DAh), as
described above, will toggle the flip-flop back to writing to the index.
ARX
Attribute Controller Index Register
9.4.1
Default:
Attributes:
7
6
Reserved (00)
Bit
7:6
5
3C0h
00UU UUUUb (U = undefined)
Read/Write
5
Video
Enable
4
0
Attribute Controller Register Index
Description
Video Enable. Note that in the VGA standard, this is called the “Palette Address Source” bit.
0 = Disable. Attribute controller color registers (AR[00:0F]) can be accessed by the CPU.
1 = Enable. Attribute controller color registers (AR[00:0F]) cannot be accessed by the CPU.
4:0
Attribute Controller Register Index. These five bits are used to select any one of the attribute controller
registers (AR[00:14]) to be accessed.
Note:
AR12 is referred to in the VGA standard as the Color Plane Enable Register. “Plane,” “color plane,”
“display memory plane,” and “memory map” have been all been used in IBM literature on the VGA
standard, in order to describe the four separate regions in the frame buffer where the pixel color or
attribute information is split up and stored in standard VGA planar modes. This use of multiple terms
for the same subject was deemed to be confusing. Therefore, AR12 is called the Memory Plane
Enable Register. Attribute Controller Register Index.
124
®
R
AR[00:0F]
Palette Registers [0:F]
9.4.2
Default:
Attributes:
7
6
Reserved
Bit
Read at 3C1h and write at 3C0h; (index=00h-0Fh)
Read/Write
5
0
Palette Bits P[5:0]
Description
7:6
5:0
Palette Bits P[5:0]. In each of the 16 registers, these are the lower 6 of 8 bits that are used to map either
text attributes or pixel color input values (for modes that use 16 colors) to the 256 possible colors
selectable in the palette.
Note:
Bits 3 and 2 of the Color Select Register (AR14) supply bits P7 and P6 for the values contained in all
16 of these registers. Bits 1 and 0 of the Color Select Register (AR14) can also replace bits P5 and P4
for the values contained in all 16 of these registers, if bit 7 of the Mode Control Register (AR10) is set
to 1.
125
®
9.4.3
R
AR10
Mode Control Register
Default:
Attributes:
Read at 3C1h and write at 3C0h; (index=10h)
UUh (U = undefined)
Read/Write
7
6
5
4
3
2
1
0
Palette
Bits P5,
P4 Select
Pixel
Width/Clk
Select
Pixel
Panning
Compat
Reserved
(0)
Enable
Blink/
Select
Bkgnd Int
Enable
Line
Graphics
Char Code
Select
Display
Type
Graphics/
Alpha
Mode
Bit
7
Description
Palette Bits P5, P4 Select.
0 = P5 and P4 for each of the 16 selected colors (for modes that use 16 colors) are individually provided by
bits 5 and 4 of their corresponding Palette Registers (AR[00:0F]).
1 = P5 and P4 for all 16 of the selected colors (for modes that use 16 colors) are provided by bits 1 and 0
of Color Select Register (AR14).
6
Pixel Width/Clock Select.
0 = Six bits of video data (translated from 4 bits via the palette) are output every dot clock.
1 = Two sets of 4 bits of data are assembled to generate 8 bits of video data that is output every other dot
clock, and the Palette Registers (AR[00:0F]) are bypassed.
Note:
This bit is set to 0 for all of the standard VGA modes, except mode 13h.
5
Pixel Panning Compatibility.
0 = Horizontally scroll both the upper and lower screen regions, as specified in the Pixel Panning Register
(AR13).
1 = Horizontally scroll only the upper screen region, as specified in the Pixel Panning Register (AR13).
Note:
This bit is applicable only when the split-screen mode is used, wherein the display area is divided into
distinct upper and lower regions that function somewhat like separate displays.
4
3
Enable Blinking/Select Background Intensity.
0 = Disables blinking in graphics modes. For text modes, sets bit 7 of the character attribute bytes to
control background intensity, instead of blinking.
1 = Enables blinking in graphics modes. For text modes, sets bit 7 of the character attribute bytes to control
blinking, instead of background intensity.
Note:
The blinking rate is derived by dividing the VSYNC signal. The Blink Rate Control Register (CR82)
defines the blinking rate.
126
®
R
Bit
2
Description
Enable Line Graphics Character Code.
0 = Every 9th pixel of a horizontal line (i.e., the last pixel of each horizontal line of each 9-pixel-wide
character box) is assigned the same attributes as the background of the character of which the given
pixel is a part.
1 = Every 9th pixel of a horizontal line (i.e., the last pixel of each horizontal line of each 9-pixel-wide
character box) is assigned the same attributes as the 8th pixel of the character of which the given pixel
is a part. This setting is intended to accommodate the line-drawing characters of the PC’s extended
ASCII character set (i.e., characters with an extended ASCII code in the range B0h to DFh).
Note:
In IBM literature describing the VGA standard, the range of extended ASCII codes that are said to
include the line-drawing characters is mistakenly specified as C0h to DFh, rather than the correct range
B0h to DFh.
1
Select Display Type.
0 = Attribute bytes in text modes are interpreted as they would be for a color display.
1 = Attribute bytes in text modes are interpreted as they would be for a monochrome display.
0
Graphics/Alphanumeric Mode.
0 = Alphanumeric (text) mode.
1 = Graphics mode.
9.4.4
AR11
Overscan Color Register
Default:
Attributes:
Bit
7:0
UUh (U = undefined)
Read/Write
Description
Overscan. These 8 bits select the overscan (border) color. The border color is displayed during the
blanking intervals. For monochrome displays, this value should be set to 00h.
127
®
R
AR12
Memory Plane Enable Register
9.4.5
Default:
Attributes:
7
6
Reserved (00)
Read/Write
5
4
Video Status Mux
Bit
3
2
1
0
Enable
Plane 3
Enable
Plane 2
Enable
Plane 1
Enable
Plane 0
Description
7:6
5:4
Video Status Mux. These 2 bits are used to select 2 of the 8 possible palette bits (P7-P0) to be made
readable via bits 5 and 4 of the Input Status Register 1 (ST01). The table below shows the possible choices.
Bit [5:4]
ST01 Bit 5
ST01 Bit 4
00
01
10
11
P2 (default)
P5
P3
P7
P0 (default)
P4
P1
P6
These bits are typically unused by current software and are provided for EGA compatibility.
3:0
Enable Plane [3:0]. These 4 bits individually enable the use of each of the 4 memory planes, in providing 1
of the 4 bits used in video output to select 1 of 16 possible colors from the palette to be displayed.
0 = Disable the use of the corresponding memory plane in video output to select colors, forcing the bit that
the corresponding memory plane would have provided to a value of 0.
1 = Enable the use of the corresponding memory plane in video output to select colors.
Note:
AR12 is referred to in the VGA standard as the Color Plane Enable Register. “Plane,” “color plane,”
“display memory plane,” and “memory map” have been all been used in IBM literature on the VGA
standard, in order to describe the four separate regions in the frame buffer that are among the pixel color
or attributes information that is split up and stored in standard VGA planar modes. This use of multiple
terms for the same subject was considered confusing. Therefore, AR12 is called the Memory Plane
Enable Register.
128
®
R
AR13
Horizontal Pixel Panning Register
9.4.6
Default:
Attributes:
0Uh (U = undefined)
Read/Write
7
4
3
Reserved (0000)
Bit
0
Horizontal Pixel Shift
Description
7:4
Reserved.
3:0
Horizontal Pixel Shift 3-0. This field holds a 4-bit value that selects the number of pixels by which the
image is shifted horizontally to the left. This function is available in both text and graphics modes.
In text modes with a 9-pixel-wide character box, the image can be shifted up to 9 pixels to the left. In text
modes with an 8-pixel-wide character box and in graphics modes other than those with 256 colors, the
image can be shifted up to 8 pixels to the left.
In standard VGA mode 13h (where bit 6 of the Mode Control Register, AR10, is set to 1 to support 256
colors), bit 0 of this register must remain set to 0, and the image may be shifted up to only 4 pixels to the
left. In this mode, the number of pixels by which the image is shifted can be controlled further by using bits
6 and 5 of the Preset Row Scan Register (CR08).
Number of Pixels Shifted
Bits [3:0]
0h
1h
2h
3h
4h
5h
6h
7h
8h
9 Pixel Text
1
2
3
4
5
6
7
8
0
8-Pixel Text
& Graphics
0
1
2
3
4
5
6
7
Undefined
256-Color
Graphics
0
Undefined
1
Undefined
2
Undefined
3
Undefined
Undefined
129
®
R
AR14
Color Select Register
9.4.7
Default:
Attributes:
0Uh (U = undefined)
Read/Write
7
4
Reserved (0000)
Bit
3
2
1
0
P7
P6
Alt P5
Alt P4
Description
7:4
Reserved.
3:2
Palette Bits P[7:6]. These are the 2 uppermost of the 8 bits used to map either text attributes or pixel
color input values (for modes that use 16 colors) to the 256 possible colors contained in the palette. These
2 bits are common to all 16 sets of bits P5 through P0, that are individually supplied by Palette Registers 0F (AR[00:0F]).
1:0
Alternate Palette Bits P[5:4]. These 2 bits can be used as an alternate version of palette bits P5 and P4.
Unlike the P5 and P4 bits that are individually supplied by Palette Registers 0-F (AR[00:0F]), these 2
alternate palette bits are common to all 16 of Palette Registers. Bit 7 of the Mode Control Register (AR10)
is used to select between the use of either the P5 and P4 bits that are individually supplied by the 16
Palette Registers or these 2 alternate palette bits.
130
R
9.5
®
VGA Color Palette Registers
The palette DAC has two main components: a palette in which a selection of 256 colors may be stored as well as a set of
three digital to analog (D-to-A) converters, one each for the red, green, and blue components used to produce a color on
a CRT display. The palette DAC also is frequently called the RAMDAC, to emphasize the presence of memory
alongside the three D-to-A converters, and the palette itself often is referred to as the CLUT (i.e., color look-up table).
During normal use, the palette DAC is operated either in direct-color mode or indexed-color mode. Direct-color mode is
used with pixel depths of 15, 16 or 24 bits per pixel. In direct-color mode, the pixel data received from the frame buffer,
through the sequencer and the attribute controller, directly specifies the color for a given pixel. This pixel data is
preformatted such that certain bits of the pixel data for each pixel are used to provide the red, green, and blue output
values for each of the three corresponding 8-bit, D-to-A converters. The indexed-color mode is used with pixel depths of
8 bits per pixel or less. In the indexed-color mode, the incoming pixel data for each pixel is actually an 8-bit index that is
used to choose one of the 256 color data positions within the palette. Each color data position holds a 24-bit color value
that specifies the actual 8-bit red, green, and blue values for each of the three corresponding 8-bit, D-to-A converters. In
essence, the colors for each pixel are specified indirectly, with the actual choice of colors taking place in the color data
positions of the palette, while the incoming pixel data chooses from among these color data positions. This method
allows the full range of over 16 million possible colors to be accessible in modes with only 8 or fewer bits per pixel.
The color data stored in these 256 color data positions can be accessed only through a complex sub-addressing scheme,
using a data register and two index registers. The Palette Data Register at I/O address 3C9h (or memory address offset
3C1h) is the data port. The Palette Read Index Register at I/O address 3C7h (or memory address offset 3C7h) and the
Palette Write Index Register at I/O address 3C8h (or memory address offset 3C8h) are the two index registers. The
Palette Read Index Register is used to choose the color data position to be read from via the data port, while the Palette
Write Index Register is used to choose the color data position to be written to through the same data port. This
arrangement allows the same data port to be used for reading from and writing to two different color data positions.
Reading and writing the color data at a color data position involves three successive reads or writes, since the color data
stored at each color data position consists of three bytes.
To read a color data position, the index of the desired color data position must first be written to the Palette Read Index
Register. Then all three bytes of data in a given color data position may be read at the Palette Data Register. The first
byte read from the Palette Data Register retrieves the 8-bit value specifying the intensity of the red color component,
while the second and third bytes read are the corresponding 8-bit values for the green and blue color components,
respectively. After the third read operation is completed, the Palette Read Index Register is incremented automatically,
so that the data of the next color data position becomes readable. This allows the contents of all 256 color data positions
of the palette to be read by specifying only the index of the 0th color data position in the Palette Read Index Register,
and by then simply performing 768 successive reads from the Palette Data Register.
Writing a color data position entails a very similar procedure. The index of the desired color data position must first be
written to the Palette Write Index Register. Then all three bytes of data to specify a given color may be written to the
Palette Data Register. The first, second, and third bytes written to the Palette Data Register specify the intensity of the
red, green, and blue color components, respectively. One important detail is that all three of these bytes must be written
before the hardware will actually update these three values at the given color data position. After all three bytes have
been written, the Palette Write Index Register is incremented automatically, so that the data for the next color data
position becomes writeable. This allows the contents of all 256 color data positions of the palette to be written by
specifying only the index of the 0th color data position in the Palette Write Index Register, and by then simply
performing 768 successive writes to the Palette Data Register.
In addition to the standard set of 256 color data positions of the palette, there is also an alternate set of 8 color data
positions used to specify the colors used to draw the cursor, and these also are accessed using the very same subaddressing scheme. A bit in the Pixel Pipeline Configuration Register (PIXCONF) determines whether the standard 256color data positions or the alternate 8 color data positions are to be accessed through this sub-addressing scheme.
131
®
R
DACMASK
Pixel Data Mask Register
9.5.1
Default:
Attributes:
3C6h
Undefined
Read/Write
Bit
Description
Pixel Data Mask. In indexed-color mode, the 8 bits of this register are logically ANDed with the 8 bits of
pixel data received from the frame buffer for each pixel. The result of this ANDing process becomes the
actual index used to select color data positions within the palette. This has the effect of limiting the choice
of color data positions that may be specified by the incoming 8-bit data.
7:0
0 = Corresponding bit in the resulting 8-bit index is being forced to 0.
1 = Allows the corresponding bit in the resulting index to reflect the actual value of the corresponding bit in
the incoming 8-bit pixel data.
In direct-color mode, the palette is not used, and the data in this register is ignored.
DACSTATE
DAC State Register
9.5.2
Default:
Attributes:
3C7h
00h
Read Only
7
2
Reserved (000000)
Bit
1
0
DAC State
Description
7:2
1:0
DAC State. This field indicates which of the two index registers was most recently written.
132
Bits [1:0]
Index Register Indicated
00
01
10
11
Palette Write Index Register at I/O Address 3C7h (default)
Reserved
Reserved
Palette Read Index Register at I/O Address 3C8h
®
R
9.5.3
DACRX
Palette Read Index Register
Default:
Attributes:
3C7h
00h
Write Only
Bit
7:0
9.5.4
Description
Palette Read Index. The 8-bit index value programmed into this register selects which of 256 standard
color data positions within the palette (or which of 8 alternate color data positions, depending on the state
of a bit in the Pixel Pipeline Control 0 Register) are to be made readable via the Palette Data Register
(DACDATA). The index value held in this register is incremented automatically after all three bytes of the
color data position selected by the current index have been read.
DACWX
Palette Write Index Register
Default:
Attributes:
3C8h
00h
Write Only
Bit
7:0
9.5.5
Description
Palette Write Index. The 8-bit index value programmed into this register selects which of 256 standard
color data positions within the palette (or which of 8 alternate color data positions, depending on the state
of a bit in the Pixel Pipeline Control 0 Register) are to be made writeable via the Palette Data Register
(DACDATA). The index value held in this register is incremented automatically after all three bytes of the
color data position selected by the current index have been written.
DACDATA
Palette Data Register
Default:
Attributes:
Bit
7:0
3C9h
Undefined
Read/Write
Description
Palette Data. This byte-wide data port provides read or write access to the three bytes of data of each
color data position selected using the Palette Read Index Register (DACRX) or the Palette Write Index
Register (DACWX).
The three bytes in each color data position are read or written in three successive read or write operations.
The first, second, and third bytes read or written specify the intensity of the red, green, and blue
components, respectively, of the color specified at the selected color data position. When writing data to a
color data position, all three bytes must be written before the hardware will actually update the three bytes
of the selected color data position.
When reading or writing to a color data position, ensure that neither the Palette Read Index Register
(DACRX) nor the Palette Write Index Register (DACWX) is written to before all three bytes are read or
written. A write to either of these two registers causes the circuitry that automatically cycles to provide
access to the bytes for the red, green, and blue components to be reset in such a manner that that the byte
for the red component is the byte accessed by the next read or write operation, via this register.
133
®
9.6
R
CRT Controller Register
The CRT controller registers are accessed by writing the index of the desired register into the CRT Controller Index
Register at I/O address 3B4h or 3D4h, depending on whether the graphics system is configured for MDA or CGA
emulation. The desired register is then accessed through the data port for the CRT controller registers located at I/O
address 3B5h or 3D5h, again depending upon the choice of MDA or CGA emulation, as per MSR[0]. For memorymapped accesses, the Index register is at 3B4h (MDA mode) or 3D3h (CGA mode), and the data port is accessed at
3B5h (MDA mode) or 3D5h (CGA mode).
Notes:
Register CR80 enables / disables the CRTC extensions.
Group 0 Protection: In the original IBM VGA, CR[0:7] could be write-protected by means of CR11[7]. In BIOS
code, this write protection is set following each mode change. Other protection groups have no current use and
will not be used in the future by the BIOS or drivers. They are the result of an industry fad some years ago to
attempt to write-protect other groups of registers. However, all such schemes were chip specific. Only the IBMcompatible write protection (Group 0 Protection) is supported.
The following figure shows the display fields and dimensions as well as the particular CRxx register that provides the
control:
CR00 Horizontal Total
CR03 Horizontal Blank End
CR05 Horizontal Sync End
CR04 Horizontal Sync Start
CR02 Horizontal Blank Start
Horizontal Sync
Display
Horizontal Right Border
CR15 Vertical Blank Start
CR12 Vertical Display Enable End
CR10 Vertical Sync Start
CR11 Vertical Sync End
CR06 Vertical Total
CR16 Vertical Blank End
CR01 Horizontal Display Enable End
Vertical Bottom Border
Blank (Vertical Front Porch)
Vertical Sync
Blank (Vertical Back Porch)
Vertical Top Border
crt_reg.vsd
134
®
R
9.6.1
CRX
CRT Controller Index Register
Default:
Attributes:
3B4h/3D4h
0Uh (U = undefined)
Read/Write
Bit
Description
7:0
CRT Controller Register Index. These 8 bits are used to select any one of the CRT controller registers to
be accessed via the data port at I/O location 3B5h or 3D5h, depending upon whether the graphics system
is configured for MDA or CGA emulation. The data port memory address offsets are 3B5h/3D5h.
9.6.2
CR00
Horizontal Total Register
Default:
Attributes:
00h
Read/Write (Group 0 Protection)
This register is used to specify the total length of each scan line. This encompasses both the part of the scan line that is
within the active display area as well as the part that is outside of it. This register is extended to cover 16×12 resolution
using CR35 and CR39.
Bit
7:0
9.6.3
Description
Horizontal Total. This field should be programmed with a value equal to the total number of character
clocks within the entire length of a scan line, minus 5.
CR01
Horizontal Display Enable End Register
Default:
Attributes:
Undefined
This register is used to specify the end of the part of the scan line that is within the active display area, relative to its
beginning. In other words, this is the horizontal width of the active display area.
Bit
7:0
Description
Horizontal Display Enable End. This field should be programmed with a value equal to the number of
character clocks that occur within the part of a scan line that is within the active display area, minus 1.
135
®
R
CR02
Horizontal Blanking Start Register
9.6.4
Default:
Attributes:
Undefined
This register is used to specify the beginning of the horizontal blanking period, relative to the beginning of the active
display area of a scan line.
Bit
Description
7:0
Horizontal Blanking Start. This field should be programmed with a value equal to the number of character
clocks that occur on a scan line, from the beginning of the active display area to the beginning of the
horizontal blanking.
CR03
Horizontal Blanking End Register
9.6.5
Default:
Attributes:
7
1UUU UUUUb (U = undefined)
6
Reserved
(0)
5
4
Display Enable Skew
Control
Bit
7
6:5
4:0
0
Horizontal Blanking End Bits 4:0
Description
Reserved. Values written to this bit are ignored, and to maintain consistency with the VGA standard, a
value of 1 is returned when this bit is read. At one time, this bit was used to enable access to certain light
pen registers. At that time, setting this bit to 0 provided this access, but setting this bit to 1 was necessary
for normal operation.
Display Enable Skew Control. Defines the degree to which the start and end of the active display area
are delayed along the length of a scan line, in order to compensate for internal pipeline delays. These 2 bits
describe the delay in terms of the number character clocks.
Bit [6:5]
Amount of Delay
00
01
10
11
No delay
Delayed by 1 character clock
Delayed by 2 character clocks
Horizontal Blanking End Bits [4:0]. This field provides the 5 least-significant bits of a 6-bit value that
specifies the end of the blanking period relative to its beginning on a single scan line. Bit 7 of the Horizontal
Sync End Register (CR05) supplies the most-significant bit.
This 6-bit value should be programmed to be equal to the least-significant 6 bits of the result obtained by
adding the length of the blanking period (in terms of character clocks) to the value specified in the
Horizontal Blanking Start Register (CR02).
136
®
R
9.6.6
CR04
Horizontal Sync Start Register
Default:
Attributes:
Undefined
This register is used to specify the beginning of the horizontal sync pulse relative to the beginning of the active display
area on a scan line.
Bit
7:0
9.6.7
Description
Horizontal Sync Start. This field should be set equal to the number of character clocks that occur from
beginning of the active display area to the beginning of the horizontal sync pulse on a single scan line.
CR05
Horizontal Sync End Register
Default:
Attributes:
7
Hor Blank
End Bit 5
00h
6
5
4
0
Horizontal Sync Delay
Horizontal Sync End
Bit
Description
7
Horizontal Blanking End Bit 5. This bit provides the most-significant bit of a 6-bit value that specifies the
end of the horizontal blanking period relative to its beginning. Bits [4:0] of Horizontal Blanking End Register
(CR03) supply the 5 least-significant bits. See CR03[4:0] for further details.
This 6-bit value should be set to the least-significant 6 bits of the result obtained by adding the length of the
blanking period (in terms of character clocks) to the value specified in the Horizontal Blanking Start
Register (CR02).
6:5
4:0
Horizontal Sync Delay. This field defines the degree to which the start and end of the horizontal sync
pulse are delayed in order to compensate for internal pipeline delays. This capability is supplied to
implement VGA compatibility. These field describes the delay in terms of a number character clocks.
Bit [6:5]
Amount of Delay
00
01
10
11
No delay
Horizontal Sync End.
This field provides the 5 least-significant bits of a 5-bit value that specifies the end of the horizontal sync
pulse relative to its beginning, a value equal to the 5 least-significant bits of the horizontal character counter
value, at which time the horizontal retrace signal becomes inactive (logical 0). Thus, this 5-bit value
specifies the width of the horizontal sync pulse.
To obtain a retrace signal of W, the following algorithm is used:
Value of Horizontal Sync Start Register (CR04) + width of horizontal retrace signal (in character clock units)
= 5-bit result to be programmed in this field
137
®
9.6.8
R
CR06
Vertical Total Register
Default:
Attributes:
00h
Bit
Description
7:0
Vertical Total Bits [7:0]. This field provides the 8 least-significant bits of either a 10-bit or 12-bit value that
specifies the total number of scan lines. This includes the scan lines both inside and outside of the active
display area.
specified with a 10-bit value. The 8 least-significant bits of this value are supplied by these 8 bits of this
register, and the 2 most-significant bits are supplied by bits 5 and 0 of the Overflow Register (CR07).
with a 12-bit value. The 8 least-significant bits of this value are supplied by these 8 bits of this register, and
the 4 most-significant bits are supplied by bits [3:0] of the Extended Vertical Total Register (CR30).
This 10-bit or 12-bit value should be programmed to equal the total number of scan lines, minus 2.
9.6.9
CR07
Overflow Register
Default:
Attributes:
Bit
7
UU0U UUU0b (U = undefined)
Read/Write (Group 0 Protection on bits [7:5, 3:0])
7
6
5
4
3
2
1
0
Vert Sync
Start Bit 9
Vert Disp
En Bit 9
Vert Total
Bit 9
Line Cmp
Bit 8
Vert Blnk
Start Bit 8
Vert Sync
Start Bit 8
Vert Disp
En Bit 8
Vert Total
Bit 8
Description
Vertical Sync Start Bit 9. The vertical sync start is a 10-bit or 12-bit value that specifies the beginning of
the vertical sync pulse, relative to the beginning of the active display area.
Vertical Sync Start Register (CR10), and the most-significant and second-most-significant bits are supplied
by this bit and bit 2, respectively, of this register.
Vertical Sync Start Register (CR10), and the 4 most-significant bits are supplied by bits [3:0] of the
Extended Vertical Sync Start Register (CR32) register. In extended modes, neither this bit nor bit 2 of this
register is used.
This 10-bit or 12-bit value should be programmed to be equal to the number of scan lines, from the
138
®
R
Bit
Description
6
Vertical Display Enable End Bit 9. The vertical display enable end is a 10-bit or 12-bit value that specifies
enable end is specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits [7:0]
of the Vertical Display Enable End Register (CR12), and the most-significant and second-most-significant
bits are supplied by this bit and bit 1, respectively, of this register.
Vertical Display Enable End Register (CR12), and the 4 most-significant bits are supplied by bits [3:0] of
the Extended Vertical Display End Enable Register (CR31). In extended modes, neither this bit nor bit 1 of
this register is used.
5
Vertical Total Bit 9. The vertical total is a 10-bit or 12-bit value that specifies the total number of scan
lines. This includes the scan lines both inside and outside of the active display area.
Vertical Total Register (CR06), and the most-significant and second-most-significant bits are supplied by
this bit and bit 0, respectively, of this register.
Register (CR06), and the 4 most-significant bits are supplied by [3:0] bits of the Extended Vertical Total
Register (CR30). In extended modes, neither this bit nor bit 0 of this register is used.
This 10-bit or 12-bit value should be programmed equal to the total number of scan lines, minus 2.
4
Line Compare Bit 8. This bit provides the second-most-significant bit of a 10-bit value that specifies the
scan line at which the memory address counter restarts at the value of 0. Bit 6 of the Maximum Scan Line
Register (CR09) supplies the most-significant bit, and bits 7-0 of the Line Compare Register (CR18) supply
the 8 least-significant bits.
Normally, this 10-bit value is set to specify a scan line after the last scan line of the active display area.
When this 10-bit value is set to specify a scan line within the active display area, it causes that scan line
and all subsequent scan lines in the active display area to display video data, starting at the very first byte
of the frame buffer. The result is what appears to be a screen split into a top and bottom part, with the
image in the top part being repeated in the bottom part.
When it is used in cooperation with the Start Address High Register (CR0C) and the Start Address Low
Register (CR0D), it is possible to create a split display, as described earlier, but with the top and bottom
parts displaying different data. The top part will display the data in the frame buffer, starting at the address
specified in the two aforementioned start address registers, while the bottom part will display the data in the
frame buffer, starting at the first byte of the frame buffer.
139
®
Bit
3
R
Description
Vertical Blanking Start Bit 8. The vertical blanking start is a 10-bit or 12-bit value that specifies the
beginning of the vertical blanking period, relative to the beginning of the active display area.
supplied by bit 5 of the Maximum Scan Line Register (CR09) and this bit of this register, respectively.
Vertical Blanking Start Register (CR15), and the 4 most-significant bits are supplied by bits [3:0] of the
Extended Vertical Blanking Start Register (CR33). In extended modes, neither this bit nor bit 5 of the
Maximum Scan Line Register (CR09) is used.
2
Vertical Sync Start Bit 8. The vertical sync start is a 10-bit or 12-bit value that specifies the beginning of
the vertical sync pulse, relative to the beginning of the active display area.
Vertical Sync Start Register (CR10), and the most-significant and second-most-significant bits are supplied
by bit 7 and this bit, respectively, of this register.
Vertical Sync Start Register (CR10), and the 4 most-significant bits are supplied by bits [3:0] of the
Extended Vertical Sync Start Register (CR32) register. In extended modes, neither this bit nor bit 7 of this
register is used.
1
Vertical Display Enable End Bit 8. The vertical display enable end is a 10-bit or 12-bit value that specifies
enable end is specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits [7:0]
of the Vertical Display Enable End Register (CR12), and the most-significant and second-most-significant
bits are supplied by bit 6 and this bit, respectively, of this register.
Vertical Display Enable End Register (CR12), and the 4 most-significant bits are supplied by bits [3:0] of
the Extended Vertical Display End Enable Register (CR31). In extended modes, neither this bit nor bit 6 of
this register is used.
140
®
R
Bit
0
Description
Vertical Total Bit 8. The vertical total is a 10-bit or 12-bit value that specifies the total number of scan
lines. This includes the scan lines both inside and outside of the active display area.
Vertical Total Register (CR06), and the most-significant and second-most-significant bits are supplied by bit
5 and this bit, respectively, of this register.
Register (CR06), and the 4 most-significant bits are supplied by 3-0 bits of the Extended Vertical Total
Register (CR30). In extended modes, neither this bit nor bit 5 of this register is used.
141
®
9.6.10
R
CR08
Preset Row Scan Register
Default:
Attributes:
7
Reserved
(0)
Read/Write
6
5
4
0
Byte Panning
Starting Row Scan Count
Bit
7
6:5
Description
Byte Panning. This field holds a 2-bit value that selects number of bytes (up to 3) by which the image is
shifted horizontally to the left on the screen. This function is available in both text and graphics modes.
In text modes with a 9-pixel-wide character box, the image can be shifted up to 27 pixels to the left, in
increments of 9 pixels. In text modes with an 8-pixel-wide character box and in all standard VGA graphics
modes, the image can be shifted up to 24 pixels to the left, in increments of 8 pixels.
The image can be shifted still further, in increments of individual pixels, through the use of bits [3:0] of the
Horizontal Pixel Panning Register (AR13).
Number of Pixels Shifted
Bit [6:5]
00
01
10
11
4:0
142
9-Pixel Text
0
9
18
27
8-Pixel Text & Graphics
0
8
16
24
Starting Row Scan Count. This field specifies which horizontal line of pixels within the character boxes of
the characters used on the top row of text on the display will be used as the top scan line. The horizontal
lines of pixels of a character box are numbered from top to bottom, with the top line of pixels being number
0. If a horizontal line of the these character boxes other than the top line is specified, then the horizontal
lines of the character box above the specified line of the character box will not be displayed as part of the
top row of text characters on the display. Normally, the value specified by these 5 bits should be 0, so that
all of the horizontal lines of pixels within these character boxes will be displayed in the top row of text,
ensuring that the characters in the top row of text do not appear to be cut off at the top.
®
R
9.6.11
CR09
Maximum Scan Line Register
Default:
Attributes:
00h
Read/Write
7
6
5
Double
Scanning
Line Cmp
Bit 9
Vert Blnk
Start Bit 9
Bit
7
4
0
Starting Row Scan Count
Description
Double Scanning Enable.
0 = Disable. When disabled, the clock to the row scan counter is equal to the horizontal scan rate. This is
the normal setting for many of the standard VGA modes and all of the extended modes.
1 = Enable. When enabled, the clock to the row scan counter is divided by 2. This is normally used to allow
CGA-compatible modes that have only 200 scan lines of active video data to be displayed as 400 scan
lines (by displaying each scan line twice).
6
Line Compare Bit 9. This bit provides the most-significant bit of the 10-bit value that specifies the scan line
at which the memory address counter restarts at the value of 0. Bit 4 of the Overflow Register (CR07)
supplies the second-most-significant bit, and bits 7-0 of the Line Compare Register (CR18) supply the 8
least-significant bits.
of the frame buffer. The result is what appears to be a screen split into top and bottom parts, with the image
in the top part being repeated in the bottom part.
parts displaying different data. The top part will display whatever data exists in the frame buffer starting at
the address specified in the two aforementioned start address registers, while the bottom part will display
whatever data exists in the frame buffer starting at the first byte of the frame buffer.
5
Vertical Blanking Start Bit 9. The vertical blanking start is a 10-bit or 12-bit value that specifies the
beginning of the vertical blanking period relative to the beginning of the active display area.
supplied by this bit and bit 3 of the Overflow Register (CR09), respectively.
Vertical Blanking Start Register (CR15), and the 4 most-significant bits are supplied by bits [3:0] of the
Extended Vertical Blanking Start Register (CR33). In extended modes, neither this bit nor bit 3 of the
Overflow Register (CR09) is used.
4:0
Starting Row Scan Count. This field provides all 5 bits of a 5-bit value that specifies the number of scan
lines in a horizontal row of text. This value should be programmed to be equal to the number of scan lines in
a horizontal row of text, minus 1.
143
®
9.6.12
R
CR0A
Text Cursor Start Register
Default:
Attributes:
3B5h/3D5h (index=0Ah)
Read/Write
This cursor is the text cursor that is part of the VGA standard and should not be confused with the hardware cursor and
pop-up (a.k.a., cursor and cursor 2), which are intended to be used in graphics modes. This text cursor exists only in text
modes. Therefore, this register is entirely ignored in graphics modes.
7
6
Reserved (00)
5
5
0
Text
Cursor Off
Bit
7:6
4
Text Cursor Start
Description
Text Cursor Off.
0 = Enables the text cursor.
1 = Disables the text cursor.
4:0
144
Text Cursor Start. This field specifies the horizontal line of pixels in a character box that is to be used in
order to display the first horizontal line of the cursor in text mode. The horizontal lines of pixels in a
character box are numbered from top to bottom, with the top line being number 0. The value specified by
these 5 bits should be the number of the first horizontal line of pixels on which the cursor is to be shown.
®
R
9.6.13
CR0B
Text Cursor End Register
Default:
Attributes:
3B5h/3D5h (index=0Bh)
Read/Write
modes, so this register is entirely ignored in graphics modes.
7
6
Reserved
(0)
5
4
Text Cursor Skew
Bit
7
6:5
4:0
3
2
1
0
Text Cursor End
Description
Text Cursor Skew. This field specifies the degree to which the start and end of each horizontal line of
pixels making up the cursor is delayed, in order to compensate for internal pipeline delays. These 2 bits
describe the delay in terms of a number character clocks.
Bit [6:5]
Amount of Delay
00
01
10
11
No delay
Text Cursor End. This field specifies the horizontal line of pixels in a character box that is to be used to
display the last horizontal line of the cursor in text mode. The horizontal lines of pixels in a character box
are numbered from top to bottom, with the top line being number 0. The value specified by these 5 bits
should be the number of the last horizontal line of pixels on which the cursor is to be shown.
145
®
9.6.14
R
CR0C
Start Address High Register
Default:
Attributes:
3B5h/3D5h (index=0Ch)
Undefined
Read/Write
Bit
Description
7:0
Start Address Bits [15:8] or [17:10]. This register provides either bits 15 through 8 of the 16-bit value that
specifies the memory address offset from the beginning of the frame buffer, or bits 17 through 10 of the 32-bit
buffer address at which begins the data to be shown in the active display area. (Default: 0)
specified with a 16-bit value. The eight bits of this register provide the eight most-significant bits of this
value, while the eight bits of the Start Address Low Register (CR0D) provide the eight least-significant bits.
with a 32-bit value. Bits 31 through 24 of this value are provided by the Extended Start Address High
Register (CR42). Bits 23 through 18 of this value are provided by bits 5 through 0 of the Extended Start
Address Register (CR40). Bits 17 through 10 of this value are provided by this register. Bits 9 through 2 of
this value are provided by the Start Address Low Register (CR0D). Bits 1 and 0 of this value are always 0,
and therefore not provided. It should be further noted that, in extended modes, these 32 bits from these
four registers are double-buffered and synchronized to VSYNC, in order to ensure the smooth or
instantaneous appearance of changes that occur on the screen as a result of changes in the start address.
To change the start address in extended modes, all four registers must be set for the new value, and then
bit 7 of the Extended Start Address Register (CR40) must be set to 1. Only after this is done will the
hardware update the start address on the next VSYNC. When this update has been performed, the
hardware will set bit 7 of the Extended Start Address Register (CR40) back to 0.
146
®
R
9.6.15
CR0D
Start Address Low Register
Default:
Attributes:
Bit
7:0
3B5h/3D5h (index=0Dh)
Undefined
Read/Write
Description
Start Address Bits [7:0] or [9:2]. This register provides either bits 7 through 0 of a 16 bit value that
specifies the memory address offset from the beginning of the frame buffer, or bits 9 through 2 of a 32 bit
buffer address at which begins the data to be shown in the active display area. (Default: 0)
most-significant bits of this value, while the eight bits of this register provide the eight least-significant bits.
with a 32-bit value. Bits 31 through 24 of this value are provided by the Extended Start Address High
Register (CR42). Bits 23 through 18 of this value are provided by bits 5 through 0 of the Extended Start
Address Register (CR40). Bits 17 through 10 of this value are provided by the Start Address High Register
(CR0C). Bits 9 through 2 of this value are provided by this register. Bits 1 and 0 of this value are always 0,
and therefore not provided. It should be further noted that, in extended modes, these 32 bits from these
four registers are double-buffered and synchronized to VSYNC, in order to ensure the smooth or
instantaneous appearance of changes that occur on the screen as a result of changes in the start address.
To change the start address in extended modes, all three registers must be set for the new value, and then
bit 7 of the Extended Start Address Register (CR40) must be set to 1. Only after this is done will the
hardware update the start address on the next VSYNC. When this update has been performed, the
hardware will set bit 7 of the Extended Start Address Register (CR40) back to 0.
9.6.16
CR0E
Text Cursor Location High Register
Default:
Attributes:
3B5h/3D5h (index=0Eh)
Undefined
Read/Write
This cursor is the text cursor that is part of the VGA standard, and should not be confused with the hardware cursor and
modes, so this register is entirely ignored in graphics modes.
Bit
7:0
Description
Text Cursor Location Bits [15:8]. This field provides the 8 most-significant bits of a 16-bit value that
specifies the address offset from the beginning of the frame buffer at which the text cursor is located. Bit
7:0 of the Text Cursor Location Low Register (CR0F) provide the 8 least-significant bits.
147
®
9.6.17
R
CR0F
Text Cursor Location Low Register
Default:
Attributes:
3B5h/3D5h (index=0Fh)
Undefined
Read/Write
modes, and so this register is entirely ignored in graphics modes.
Bit
7:0
9.6.18
Description
Text Cursor Location Bits [7:0]. This field provides the 8 least-significant bits of a 16-bit value that
specifies the address offset from the beginning of the frame buffer at which the text cursor is located. Bits
7:0 of the Text Cursor Location High Register (CR0D) provide the 8 most-significant bits.
CR10
Vertical Sync Start Register
Default:
Attributes:
Bit
7:0
Undefined
Read/Write
Description
Vertical Sync Start Bits [7:0]. This register provides the 8 least-significant bits of either a 10-bit or 12-bit
value that specifies the beginning of the vertical sync pulse relative to the beginning of the active display
area of a screen.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, this value is described
in 10 bits, with bits [7,2] of the Overflow Register (CR07) supplying the 2 most-significant bits.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, this value is described in 12
bits, with bits [3:0] of the Extended Vertical Sync Start Register (CR32) supplying the 4 most-significant
bits.
This 10-bit or 12-bit value should equal the vertical sync start, in terms of the number of scan lines from the
beginning of the active display area to the beginning of the vertical sync pulse. Since the active display
area always starts on the 0th scan line, this number should equal the number of the scan line on which the
148
®
R
9.6.19
CR11
Vertical Sync End Register
Default:
Attributes:
0U00 UUUUb (U = undefined)
Read/Write
7
6
5
4
Protect
Regs 0:7
Reserved
Vert Int
Enable
Vert Int
Clear
Bit
7
3
0
Vertical Sync End
Description
Protect Registers [0:7]. Note that the ability to write to bit 4 of the Overflow Register (CR07) is not
affected by this bit (i.e., bit 4 of the Overflow Register is always writeable).
0 = Enable writes to registers CR[00:07]. (default)
1 = Disable writes to registers CR[00:07].
6
Reserved. In the VGA standard, this bit was used to switch between 3 and 5 frame buffer refresh cycles,
during the time required to draw each horizontal line.
5
Vertical Interrupt Enable. Note that the graphics controller does not provide an interrupt signal that would
be connected to an input of the system’s interrupt controller. Bit 7 of Input Status Register 0 (ST00)
indicates the status of the vertical retrace interrupt, and it can be polled by software to determine whether a
vertical retrace interrupt has taken place. Bit 4 of this register can be used to clear a pending vertical
retrace interrupt.
0 = Enable the generation of an interrupt at the beginning of each vertical retrace period.
1 = Disable the generation of an interrupt at the beginning of each vertical retrace period.
4
Vertical Interrupt Clear. Note that the graphics controller does not provide an interrupt signal that would
be connected to an input of the system’s interrupt controller. Bit 7 of Input Status Register 0 (ST00)
indicates the status of the vertical retrace interrupt, and it can be polled by software to determine whether a
vertical retrace interrupt has taken place. Bit 5 of this register can be used to enable or disable the
generation of vertical retrace interrupts.
0 = Setting this bit to 0 clears a pending vertical retrace interrupt. This bit must be set back to 1 to enable
the generation of another vertical retrace interrupt.
3:0
Vertical Sync End. This 4-bit field provides a 4-bit value that specifies the end of the vertical sync pulse
relative to its beginning. This 4-bit value should be set to the least-significant 4 bits of the result obtained by
adding the length of the vertical sync pulse (in terms of the number of scan lines that occur within the
length of the vertical sync pulse) to the value that specifies the beginning of the vertical sync pulse. (For
further details, see the description of the Vertical Sync Start Register.)
149
®
9.6.20
R
CR12
Vertical Display Enable End Register
Default:
Attributes:
Undefined
Read/Write
Bit
7:0
Description
Vertical Display Enable End Bits [7:0]. This register provides the 8 least-significant bits of either a 10-bit
or 12-bit value that specifies the number of the last scan line within the active display area.
In standard VGA modes, where bit 0 of the I/O Control Register (CR80) is set to 0, this value is described
in 10 bits, with bits [6,1] of the Overflow Register (CR07) supplying the 2 most-significant bits.
In extended modes, where bit 0 of the I/O Control Register (CR80) is set to 1, this value is described in 12
bits, with bits [3:0] of the Extended Vertical Display Enable End Register (CR31) supplying the 4 mostsignificant bits.
9.6.21
CR13
Offset Register
Default:
Attributes:
Bit
7:0
Undefined
Read/Write
Description
Offset Bits [7:0]. This register provides either all 8 bits of an 8-bit value or the 8 least-significant bits of a
12-bit value that specifies the number of words or dwords of frame buffer memory occupied by each
horizontal row of characters. Whether this value is interpreted as the number of words or dwords is
determined by the settings of the bits in the Clocking Mode Register (SR01).
with an 8-bit value, all bits of which are provided by this register.
12-bit value. The four most-significant bits of this value are provided by bits [3:0] of the Extended Offset
Register, and the eight least-significant bits are provided by this register.
This 8-bit or 12-bit value should be programmed to equal either the number of words or dwords (depending
on the setting of the bits in the Clocking Mode Register, SR01) of frame buffer memory that is occupied by
each horizontal row of characters.
150
®
R
9.6.22
CR14
Underline Location Register
Default:
Attributes:
7
Reserved
(0)
Read/Write
6
5
Dword
Mode
Count By
4
Bit
4
0
Underline Location
Description
7
6
DWord Mode.
0 = Frame buffer addresses are interpreted by the frame buffer address decoder as being either byte
addresses or word addresses, depending on the setting of bit 6 of the CRT Mode Control Register
(CR17).
1 = Frame buffer addresses are interpreted by the frame buffer address decoder as being dword
addresses, regardless of the setting of bit 6 of the CRT Mode Control Register (CR17).
Note that this bit is used in conjunction with bits 6 and 5 of the CRT Mode Control Register (CR17), in order
to select how frame buffer addresses from the CPU are interpreted by the frame buffer address decoder,
as shown below:
CR14[6]
CR17[6]
Addressing Mode
0
0
Word mode
0
1
Byte mode
1
0
DWord mode
1
1
DWord mode
5
Count By 4.
0 = The memory address counter is incremented either every character clock or every other character
clock, depending upon the setting of bit 3 of the CRT Mode Control Register.
1 = The memory address counter is incremented either every 4 character clocks or every 2 character
clocks, depending upon the setting of bit 3 of the CRT Mode Control Register.
Note that this bit is used in conjunction with bit 3 of the CRT Mode Control Register (CR17), in order to
select the number of character clocks required to cause the memory address counter to be incremented, as
shown below:
CR14[5]
CR17[3]
Address Incrementing Interval
0
0
Every character clock
0
1
Every 2 character clocks
1
0
1
1
4:0
Underline Location. This field specifies which horizontal line of pixels in a character box is to be used to
display a character underline in text mode. The horizontal lines of pixels in a character box are numbered
from top to bottom, with the top line being number 0. The value specified by these 5 bits should be the
number of the horizontal line on which the character underline mark is to be shown.
151
®
9.6.23
R
CR15
Vertical Blanking Start Register
Default:
Attributes:
Undefined
Read/Write
Bit
Description
7:0
Vertical Blanking Start Bits [7:0]. This register provides the 8 least-significant bits of either a 10-bit or 12bit value that specifies the beginning of the vertical blanking period relative to the beginning of the active
display area of the screen. Whether this value is described in 10 or 12 bits depends on the setting of bit 0
of the I/O Control Register (CR80).
start is specified with a 10-bit value. The most-significant and second-most-significant bits of this value are
respectively.
specified with a 12-bit value. The 4 most-significant bits of this value are supplied by bits [3:0] of the
Extended Vertical Blanking Start Register (CR33).
This 10-bit or 12-bit value should be programmed to be equal the number of scan lines from the beginning
of the active display area to the beginning of the vertical blanking period. Since the active display area
always starts on the 0th scan line, this number should equal the number of the scan line on which vertical
blanking begins.
9.6.24
CR16
Vertical Blanking End Register
Default:
Attributes:
Undefined
Read/Write
This register provides an 8-bit value that specifies the end of the vertical blanking period, relative to its beginning.
Bit
7:0
152
Description
Vertical Blanking End Bits [7:0]. This 8-bit value should be set equal to the least-significant 8 bits of the
result obtained by adding the length of the vertical blanking period (in terms of the number of scan lines
that occur within the length of the vertical blanking period) to the value that specifies the beginning of the
vertical blanking period. (For details, see the description of the Vertical Blanking Start Register.)
®
R
9.6.25
CR17
CRT Mode Control
Default:
Attributes:
0UU0 UUUUb (U = undefined)
Read/Write
7
6
5
4
3
2
1
0
CRT Ctrl
Reset
Word or
Byte
Mode
Address
Wrap
Reserved
(0)
Count By
2
Horizontal
Retrace
Select
Select
Row Scan
Cntr
Compat
Mode
Support
Bit
7
Description
CRT Controller Reset.
0 = Forces horizontal and vertical sync signals to be inactive. No other registers or outputs are affected.
1 = Permits normal operation.
6
Word Mode or Byte Mode.
0 = The memory address counter’s output bits are shifted by 1 bit position before being passed on to the
frame buffer address decoder, in such a manner that they are made into word-aligned addresses when
bit 6 of the Underline Location Register (CR17) is set to 0.
1 = The memory address counter’s output bits remain unshifted before being passed on to the frame buffer
address decoder, in such a manner that they remain byte-aligned addresses when bit 6 of the
Underline Location Register (CR17) is set to 0.
Note that this bit is used in conjunction with bits 6 and 5 of the CRT Mode Control Register (CR17), in
order to control how frame buffer addresses from the memory address counter are interpreted by the
frame buffer address decoder, as shown below:
CR14[6]
CR17[6]
Addressing Mode
0
0
Word Mode  Addresses from the memory address counter are shifted once to
become word-aligned.
0
1
Byte Mode  Addresses from the memory address counter are not shifted.
1
0
Dword Mode  Addresses from the memory address counter are shifted twice to
become dword-aligned.
1
1
Dword Mode  Addresses from the memory address counter are shifted twice to
become dword-aligned.
5
Address Wrap. Note that this bit is effective only when the word mode is made active by setting bit 6 in
both the Underline Location Register and this register to 0.
0 = Wrap frame buffer address at 16 KB. This is used in CGA-compatible modes.
1 = No wrapping of frame buffer addresses
4
153
®
R
Bit
Description
3
Count By 2. This bit is used in conjunction with bit 5 of the Underline Location Register (CR14), in order to
select the number of character clocks required to cause the memory address counter to be incremented.
0 = The memory address counter is incremented either every character clock or every 4 character clocks,
depending upon the setting of bit 5 of the Underline Location Register.
1 = The memory address counter is incremented every other clock.
CR14[5]
0
0
1
1
2
CR17[3]
0
1
0
1
Address Incrementing Interval
Every character clock
Horizontal Retrace Select. This bit provides a way of effectively doubling the vertical resolution, by
allowing the vertical timing counter to be clocked by the horizontal retrace clock divided by 2. (Usually, it
would be undivided.)
0 = The vertical timing counter is clocked by the horizontal retrace clock.
1 = The vertical timing counter is clocked by the horizontal retrace clock divided by 2.
1
Select Row Scan Counter.
0 = A substitution takes place: Bit 14 of the 16-bit memory address generated by the memory address
counter (after the stage at which these 16 bits may have already been shifted to accommodate word or
dword addressing) is replaced with bit 1 of the row scan counter at the stage just before this address is
presented to the frame buffer address decoder.
1 = No substitution takes place. See the following tables.
0
Compatibility Mode Support.
0 = A substitution takes place: Bit 13 of the 16-bit memory address generated by the memory address
counter (after the stage at which these 16 bits may have already been shifted to accommodate word or
dword addressing) is replaced with bit 0 of the row scan counter at a stage just before this address is
presented to the frame buffer address decoder.
1 = No substitution takes place. See the following tables.
The following tables show the possible ways in which the address bits from the memory address counter can be shifted
and/or reorganized, before being presented to the frame buffer address decoder. First, the address bits generated by the
memory address counter are reorganized, if need be, in order to accommodate byte, word or dword modes. The resulting
reorganized outputs (MAOut15-MAOut0) from the memory address counter may also be further manipulated with the
substitution of bits from the row scan counter (RSOut1 and RSOut0), before finally being presented to the input bits of
the frame buffer address decoder (FBIn15-FBIn0).
154
®
R
Table 6. Memory Address Counter Address Bits [15:0]
MAOut0
MAOut1
MAOut2
MAOut3
MAOut4
MAOut5
MAOut6
MAOut7
MAOut8
MAOut9
MAOut10
MAOut11
MAOut12
MAOut13
MAOut14
MAOut15
Byte Mode
CR14 bit 6=0
CR17 bit 6=1
CR17 bit 5=X
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Word Mode
CR14 bit 6=0
CR17 bit 6=0
CR17 bit 5=1
15
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Word Mode
CR14 bit 6=0
CR17 bit 6=0
CR17 bit 5=0
13
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
DWord Mode
CR14 bit 6=1
CR17 bit 6=X
CR17 bit 5=X
12
13
0
1
2
3
4
5
6
7
8
9
10
11
12
13
CR17 bit 1=1
CR17 bit 1=1
CR17 bit 1=0
CR17 bit 1=0
CR17 bit 0=1
CR17 bit 0=0
CR17 bit 0=1
CR17 bit 0=0
MAOut0
MAOut1
MAOut2
MAOut3
MAOut4
MAOut5
MAOut6
MAOut7
MAOut8
MAOut9
MAOut10
MAOut11
MAOut12
MAOut13
MAOut14
MAOut15
MAOut0
MAOut1
MAOut2
MAOut3
MAOut4
MAOut5
MAOut6
MAOut7
MAOut8
MAOut9
MAOut10
MAOut11
MAOut12
MAOut13
RSOut1
MAOut15
MAOut0
MAOut1
MAOut2
MAOut3
MAOut4
MAOut5
MAOut6
MAOut7
MAOut8
MAOut9
MAOut10
MAOut11
MAOut12
RSOut0
MAOut14
MAOut15
MAOut0
MAOut1
MAOut2
MAOut3
MAOut4
MAOut5
MAOut6
MAOut7
MAOut8
MAOut9
MAOut10
MAOut11
MAOut12
RSOut0
RSOut1
MAOut15
X = Don’t Care
Table 7. Frame Buffer Address Decoder
FBIn0
FBIn1
FBIn2
FBIn3
FBIn4
FBIn5
FBIn6
FBIn7
FBIn8
FBIn9
FBIn10
FBIn11
FBIn12
FBIn13
FBIn14
FBIn15
155
®
9.6.26
R
CR18
Line Compare Register
Default:
Attributes:
Undefined
Read/Write
Bit
Description
7:0
Line Compare Bits [7:0]. This register provides the 8 least-significant bits of the 10-bit value that specifies
the scan line at which the memory address counter restarts at the value of 0. Bit 6 of the Maximum Scan
Line Register (CR09) supplies the most-significant bit, and bit 4 of the Overflow Register (CR07) supplies
the second-most-significant bit.
of the frame buffer. The result is what appears to be a screen split into top and bottom parts, with the
image in the top part being repeated in the bottom part. (This register is used only in split-screening
modes, which is not a problem because split screening is not actually used for extended modes. As a
result, there is no benefit to extending the existing overflow bits for higher resolutions. )
parts displaying different data. The top part will display whatever data exists in the frame buffer starting at
the address specified in the two aforementioned start address registers, while the bottom part will display
whatever data exists in the frame buffer starting at the first byte of the frame buffer.
9.6.27
CR22
Memory Read Latch Data Register
Default:
Attributes:
Bit
7:0
156
00h
Read Only
Description
Memory Read Latch Data. This field provides the value currently stored in 1 of the 4 memory read
latches. Bits 1 and 0 of the Read Map Select Register (GR04) select which of the 4 memory read latches
may be read via this register.
®
R
9.6.28
CR24
 Test Register for Toggle State of Attribute Controller Register
Default:
Attributes:
7
00h
Read Only
6
0
Toggle
Status
Reserved (0000000)
Bit
Description
Toggle Status. Last write to attribute register was to:
7
0 = index port
1 = data port
6:0
9.6.29
CR30
Extended Vertical Total Register
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Total Bits 11:8
Description
7:4
3:0
Vertical Total Bits [11:8]. The vertical total is a 10-bit or 12-bit value that specifies the total number of
scan lines. This includes the scan lines both inside and outside of the active display area.
Vertical Total Register (CR06), and the 2 most-significant bits are supplied by bits 5 and 0 of the Overflow
Register (CR07). In standard VGA modes, these 4 bits of this register are not used.
Register (CR06), and the 4 most-significant bits are supplied by these 4 bits of this register.
157
®
9.6.30
R
CR31
Extended Vertical Display End Register
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Display End Bits 11:8
Description
7:4
3:0
Vertical Display End Bits [11:8]. The vertical display enable end is a 10-bit or 12-bit value that specifies
enable end is specified with a 10-bit value. The 8 least-significant bits of this value are supplied by bits
[7:0] of the Vertical Display Enable End Register (CR12), and the 2 most-significant bits are supplied by
bits 6 and 1 of the Overflow Register (CR07). In standard VGA modes these 4 bits of this register are not
used.
Vertical Display Enable End Register (CR12), and the 4 most-significant bits are supplied by these 4 bits of
this register.
158
®
R
9.6.31
CR32
Extended Vertical Sync Start Register
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Sync Start Bits 11:8
Description
7:4
3:0
Vertical Sync Start Bits [11:8]. The vertical sync start is a 10-bit or 12-bit value that specifies the
beginning of the vertical sync pulse, relative to the beginning of the active display area.
Vertical Sync Start Register (CR10), and the 2 most-significant bits are supplied by bits 7 and 2 of the
Overflow Register (CR07). In standard VGA modes, these 4 bits of this register are not used.
Vertical Sync Start Register (CR10), and the 4 most-significant bits are supplied by these 4 bits of this
register.
159
®
9.6.32
R
CR33
Extended Vertical Blanking Start Register
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Vertical Blanking Start Bits 11:8
Description
7:4
3:0
Vertical Blanking Start Bits [11:8]. The vertical blanking start is a 10-bit or 12-bit value that specifies the
beginning of the vertical blanking period, relative to the beginning of the active display area.
respectively. In standard VGA modes, these four bits are not used.
specified with a 12-bit value. The 8 least-significant bits of this value are supplied by bits {7:0] of the
Vertical Blanking Start Register (CR15), and the 4 most-significant bits are supplied by these 4 bits of this
register.
160
®
R
9.6.33
CR34 Extended Vertical Blank Time Register
(Reserved, Not Implemented)
9.6.34
CR35
 Extended Horizontal Total Time Register
Default:
Attributes:
00h
Read/Write
7
1
Ext Horiz
Total
Reserved (0000000)
Bit
0
Description
Reserved. When this register is written to, these bits should be set to 0.
7:1
Extended Horizontal Total (MSB that extends CR00).
0
9.6.35
CR39
Extended Horizontal Blank Time Register
Default:
Attributes:
00h
Read/Write
7
1
Reserved (0000000)
Bit
0
Ext Horiz
Total
Description
Reserved.
7:1
Extended Horizontal Total (MSB that extends CR5[7], CR3[4:0]).
0
9.6.36
CR40
Extended Start Address Register
Default:
Attributes:
7
6
Start Addr
Enable
Reserved
(0)
00h
Read/Write
5
0
Start Address Bits 23:18
161
®
R
Bit
Description
7
Extended Mode Start Address Enable. This bit is used only in extended modes, where bit 0 of the I/O
Control Register (CR80) is set to 1, in order to signal the graphics controller to update the start address. In
extended modes, the start address is specified with a 30-bit value. These 30 bits, which are provided by
the Start Address Low Register (CR0D), the Start Address High Register (CR0C), the Extended Start
Address High Register (CR42), and bits [5:0] of this register, are double-buffered and synchronized to
VSYNC, in order to ensure that changes occurring on the screen as a result of changes in the start
address always have a smooth or instantaneous appearance. To change the start address in extended
modes, all three registers must be set for the new value, and then this bit of this register must be set to 1.
Only after this is done will the graphics controller update the start address on the next VSYNC. After this
update has been performed, the graphics controller will set bit 7 of this register back to 0.
6
5:0
Start Address Bits [23:18]. This start address is either the 16-bit value that specifies the memory address
offset from the beginning of the frame buffer or the 32-bit buffer address at which begins the data to be shown
in the active display area. (Default: 0)
most-significant bits of this value, while the eight bits of the Start Address Low Register (CR0D) provide
the eight least-significant bits.
with a 32-bit value. Bits 31:24 of this value are provided by the Extended Start Address High Register
(CR42). Bits 23:18 of this value are provided by bits 5:0 of this register. Bits 17:10 of this value are
provided by the Start Address High Register (CR0C). Bits 9:2 of this value are provided by the Start
Address Low Register (CR0D). Bits 1:0 of this value are always 0 and therefore not provided. Note that, in
extended modes, these 32 bits from these four registers are double-buffered and synchronized to VSYNC,
in order to ensure that changes occurring on the screen as a result of changes in the start address always
have a smooth or instantaneous appearance. To change the start address in extended modes, all four
registers must be set for the new value, and then bit 7 of this register must be set to 1. Only after this is
162
®
R
9.6.37
CR41
Extended Offset Register
Default:
Attributes:
00h
Read/Write
7
4
3
Reserved (0000)
Bit
0
Offset Bits 11:8
Description
7:4
3:0
Offset Bits [11:8]. The offset is an 8-bit or 12-bit value that describes the number of words or dwords of
frame buffer memory occupied by each horizontal row of characters. Whether this value is interpreted as
the number of words or dwords is determined by the settings of the bits in the Clocking Mode Register
(SR01).
with an 8-bit value, all the bits of which are provided by the Offset Register (CR13).
12-bit value. The four most-significant bits of this value are provided by bits [3:0] of this register, and the
eight least-significant bits are provided by the Offset Register (CR13).
This 8-bit or 12-bit value should be programmed to be equal to either the number of words or dwords
(depending on the setting of the bits in the Clocking Mode Register, SR01) of frame buffer memory that is
occupied by each horizontal row of characters.
9.6.38
CR42
Extended Start Address High Register
Default:
Attributes:
Bit
7:0
00h
Read/Write
Description
Start Address High Bits [31:24]. This register provides bits [31:24] of the 32-bit buffer address at which
the data to be shown in the active display area begins. (Default: 0)
most-significant bits of this value, while the eight bits of the CR0D register provide the eight leastsignificant bits.
with a 32-bit value. Bits [31:24] of this value are provided by this register. Bits [23:18] of this value are
provided by bits [5:0] of the Extended Start Address Register (CR40). Bits [17:10] of this value are
provided by the Start Address High Register (CR0C). Bits [9:2] of this value are provided by the Start
Address Low Register (CR0D). Bits [1:0] of this value are always 0 and therefore not provided. It should be
further noted that, in extended modes, the 30 bits from these four registers are double-buffered and
synchronized to VSYNC, in order to ensure that changes occurring on the screen as a result of changes in
the start address always have a smooth or instantaneous appearance. To change the start address in
extended modes, all four registers must be set for the new value, and then bit 7 of the Extended Start
Address Register (CR40) must be set to 1. Only then will the graphics controller update the start address
on the next VSYNC. After the update is done, the graphics controller sets bit 7 of the Extended Start
Address Register (CR40) back to 0.
163
®
9.6.39
R
CR70
Interlace Control Register
Default:
Attributes:
7
00h
Read/Write
6
0
Interlace
Enable
CRT Half-Line Value
Bit
7
Description
Interlace Enable.
0 = Selects non-interlaced CRT output (default).
1 = Selects interlaced CRT output.
6:0
9.6.40
CRT Half-Line Value. When interlaced CRT output has been selected, the value in this register specifies
the position along the length of a scan line at which the half-line vertical sync pulse occurs for the odd
frame. This half-line vertical sync pulse begins at a position between two horizontal sync pulses on the last
scan line, rather than at the position coincident with the beginning of a horizontal sync pulse at the end of a
scan line.
CR80
I/O Control
Default:
Attributes:
00h
Read/Write
7
2
Reserved (000000)
Bit
7:2
1
1
0
Attr Cntl
Ext Enbl
CRT Cntl
Int Enbl
Description
Reserved.
Attribute Controller Extensions Enable. Controls whether the attribute registers are accessed with both
index and data at 3C0h (strict VGA mode), or whether they are accessed with 3c0h as the index and 3C1h
as data. It is possible that the BIOS software or driver software might not use the non-VGA mode. Either
method should work, but it needs to be the method the software is using.
0 = Disable (i.e., strictly VGA compatible mode) (default)
1 = Enable attribute controller extensions
The index and data of the attribute controller registers are accessible at 3C0h in standard VGA. When the
attribute controller extensions are enabled, the index and data are accessible at addresses 3C0h and
3C1h, respectively.
0
CRT Controller Interpretation Enable. This bit modifies the responses/functionality to/of registers CR30
and beyond. See CR06, CR07, CR09, CR0C, CR0D, CR10, CR12, CR13, CR15, CR30, CR32, CR33,
CR40, CR41, and CR42.
0 = Registers have strict VGA Interpretation (default)
1 = Registers have extended VGA Interpretation (i.e., access to 4-G space)
164
®
R
9.6.41
CR81
Reserved
Default:
Attributes
00h
Read/Write
This register is not present in 2d.
7
0
Reserved (00000000)
Bit
7:0
9.6.42
Description
Reserved.
CR82
Blink Rate Control
Default:
88h is the VGA default because it is more visually appealing (but the standard
VGA default is 83h).
Read/Write
Attributes
The H/W default for this register does not match the VGA compatibility requirements. The BIOS must make sure to set
this register to the correct value.
7
6
5
Character blink duty
cycle
Bit
7:6
0
VSync blink rate
Description
Character Blink Duty Cycle. (Character blink also is known as attribute blink.)
10 = 50% duty cycle (power on default).
5:0
VSync Blink Rate. Controls the cursor blink rate, in terms of the number of VSyncs, on the following
basis: The programmed value must be the (actual value/2) - 1. (Default: 3)`
165
®
R
166
R
®
10. Programming Interface
The Graphics Controller (GC) contains an extensive set of registers and instructions (also referred to as “Commands”)
for controlling 2D, 3D, and video operations. This section describes the programmer’s interface to these
registers/instructions.
10.1 Reserved Bits and Software Compatibility
In many register, instruction, and memory layout descriptions, certain bits are marked as “Reserved.” When bits are
marked as reserved, it is essential for compatibility with future devices that software treat these bits as having a
futurethough unknowneffect. The behavior of reserved bits should be regarded as not only undefined, but
unpredictable. Software should follow these guidelines in dealing with reserved bits:
Do not depend on the states of any reserved bits when testing values of registers that contain such bits. Mask out the
reserved bits before testing. Do not depend on the states of any reserved bits when storing to an instruction or to a
register.
When loading a register or formatting a instruction, always load the reserved bits with the values indicated in the
documentation, if any, or reload them with the values previously read from the register.
10.2 Overview
The GC is programmed via the three basic mechanisms:
POST-Time Programming of PCI Configuration Registers
These registers are programmed once during POST of the video device. The PCI Configuration registers are not covered
in this document. For details on accessing the graphics controller’s PCI configuration space, refer to the reference
documents mentioned in the introduction of this manual.
Direct (Physical I/O and/or Memory-Mapped I/O) Access of GC Registers
Various GC functions can be controlled only via direct register access. In addition, direct register access is required to
initiate the (asynchronous) execution of GC instruction streams. This programming mechanism is “direct” and
synchronous with software execution on the CPU.
Instruction Stream DMA (via Instruction Ring Buffers)
This programming mechanism utilizes the indirect (and asynchronous) execution of GC instruction streams in order to
control certain GC functions (e.g., all 2D and 3D drawing operations). Software writes instructions into an instruction
buffer (either a ring buffer or batch buffer) and informs the GC that the instructions are ready for execution. At some
point, the instruction parser then reads the instructions via DMA and execute them.
167
®
R
10.3 GC Register Programming
Except for the PCI configuration registers, all GC registers are memory mapped. The base address of this 512-KB
memory block is programmed in the MMADR PCI Configuration register. The figure below shows the high-level
memory map of the GC registers. Note that 2D control registers (VGA and extended VGA registers) also are located at
their standard I/O locations.
Memory Space Map
(512 KB allocation)
- Cursor Registers
- Display Registers
- Pixel Pipe Registers
Offset From
Base_Reg
7FFFFh
70000h
6FFFFh
- Reserved
60000h
5FFFFh
- Video Capture Registers
- DVD Registers
50000h
4FFFFh
Blt Engine Control Status (RO)
40000h
3FFFFh
Overlay Registers1
30000h
2FFFFh
Note:
1. Some overlay registers are doublebuffered with an additional address range
in graphics memory. See the overlay register
chapter for details.
Reserved
20000h
1FFFFh
Page Table Range
10000h
0FFFFh
Reserved
07000h
06FFFh
06000h
05FFFh
Misc I/O Control Registers
05000h
04FFFh
Reserved
04000h
03FFFh
Local Memory Interface
Control Registers
I/O Space Map
(Standard graphics locations)
- Instruction Control Regs.
- Fence Table Registers
- Interrupt Control
03000h
02FFFh
01000h
00FFFh
31
00000h
19
MMADR Register
(Base Address)
reginstm.vsd
Figure 28. Graphics controller I/O and memory map
168
®
R
10.4 GC Instruction Streams
This section describes how instruction streams can be used to control GC operation and perform GC operations.
10.4.1
Instruction Use
GC instructions are used to control drawing engines and various GC functional units:
3D Instructions. 3D instructions are used to program the 3D pipeline state and perform 3D rendering operations
(including “StretchBlt” operations).
2D Instructions (BLT). These instructions are used to perform BLT operations.
Instruction Parser (IP) Instructions. The IP instructions can be used to control and synchronize the instruction
stream as well as perform various GC auxiliary functions (e.g., define graphics buffer attributes, perform
display/overlay flips, etc.)
10.4.2
Instruction Transport Overview
Instructions are not written directly to the GC. Instead, they are placed in memory by software and later read (via DMA)
by the GC's instruction parser unit. The primary mechanism used to transport instructions utilizes two ring buffers (RB):
the low-priority RB and the interrupt RB. A secondary mechanism for instruction transport utilizes batch buffers. The IP
uses a set of rules to determine the order in which instructions are executed. The following sections in this chapter
provide descriptions of the ring buffers, batch buffers, and IP rules.
10.4.3
Instruction Parser
The GC's Instruction Parser (IP) unit is responsible for the following:
•
Detecting the presence of instructions (within the ring buffers)
•
Arbitrating the execution of instruction streams
•
Reading instructions from ring buffers and batch buffers via DMA
•
Parsing the common “client” (destination) field of instructions
•
Execution of instruction parser instructions (which control IP functionality, provide synchronization
functions, and provide miscellaneous GC control functions)
•
Redirection of 2D and 3D instructions to the appropriate destination, while following drawing engine
concurrency and coherency rules
The figure below shows a high-level diagram of the GC instruction interface.
169
®
R
Low-Priority Ring
(Graphics Memory)
Batch Buffers
Instruction
Batch Buff Instr
DMA
Instruction
Instruction access and decoding
2D Instructions
DMA
FIFO
Instr
Parser
3D Instructions (3D state,
3D Primitives, STRBLT,
Interrupt Ring
(Graphics Memory)
Batch Buffers
BLT
Engine
3D
Engine
Instruction
DMA
Batch Buff Instr
Instruction
- Parser Control
(e.g., Batch Buffer Instr., NOP,
Sync ID, Flush, breakpoint )
- Memory Interface Control
(e.g., Store DWord to memory)
- Display/Overlay Control
(e.g., Front Buffer, Scan
Lines, Overlay Flip
Display
Engine
Overlay
Engine
cmd_str.vsd
Figure 29. Graphics controller instruction interface
10.4.4
Ring Buffers (RB)
The GC provides two ring buffer (RB) mechanisms through which instructions can be passed to the instruction parser.
They are referred to as the interrupt and low-priority RBs and are basically identical, except for differences in arbitration
rules and priority.
Graphics Memory
Wrap Around
Starting Addres
Buffer Low Priority Ring
Buffer
Length
Head Pointer
DMA Pointer
Tail Pointer
Starting Addres
Head Pointer
Buffer
Length
Interrupt Ring
Buffer
DMA Pointer
Tail Pointer
RingBuf.vsd
Figure 30. Ring buffers
170
®
R
10.4.4.1
A ring buffer is defined by a set of four ring buffer registers. Before an RB can be used for instruction transport, software
must program these registers. The following fields are contained within these registers:
Ring Buffer Valid: This bit controls whether the RB is included in the instruction arbitration process. Software must
program all other RB parameters before enabling an RB. An RB can be disabled and later re-enabled. Enabling or
disabling an RB does not, of itself, change any other RB register fields.
Start Address: This field points to a contiguous, 4KB-aligned, linear (e.g., not tiled) memory address region that
provides the actual instruction buffer area.
Buffer Length: The size of the buffer, in 4-KB increments, up to 2MB
Head Offset: This is the dword offset (from the start address) of the next instruction that the IP will execute. The IP
updates this field as instructions are retired. (Note that, if instructions are pending execution, the IP will likely have
fetched instructions past the head offset). As the GC does not “reset” the head offset when an RB is enabled, the
software must program the head offset field before enabling the ring buffer. Although this allows software to enable
an RB with any legal values for head/tail (i.e., it can enable or re-enable the RB with instructions already pending), it
is anticipated that software will initialize the head offset to 0. Once the head offset reaches the tail offset (i.e., head =
tail), the IP considers the RB “empty.”
Head Wrap Count: This field is incremented by the IP every time the head offset wraps back to the start of the buffer.
As it is included in the dword written in the “report head” process, the software can use this field to track IP progress
as if the RB had a “virtual” length of 2048 times the size of the actual physical buffer.
Tail Offset: This is the qword offset (from the start Address) where software will write the next instruction. After
writing instructions into the RB, the software updates the tail offset field in order to submit the instructions for
execution (by setting it to the qword offset immediately following the last instruction to be submitted). The submitted
instructions can wrap from the end of the buffer back to the top, in which case the tail offset written will be less than
the previous value. Note that, because the RB empty condition is defined as “head offset == tail offset”, the software
must leave at least one qword free at all times. (That is, the buffer is considered “full” when only one qword is free.)
Automatic Report Head Enable: Software can request that the contents of the hardware head pointer register be
written (“reported”) to snooped system memory on a periodic basis. This is desirable because the software must use
the head offset to determine the amount of free space in the RB. Also, having the head pointer reported periodically
to system memory provides a fairly accurate head offset value automatically (i.e., without having to explicitly store a
head offset value via an instruction). The head pointer register will be stored at an RB-specific displacement into the
“hardware status page” (defined by the HWSTAM register).
171
®
R
Table 8. Ring Buffer Characteristics
Characteristic
Description
Alignment
4-KB page aligned
Max. size
2 MB
Length
Programmable in numbers of 4-KB pages
Start pointer
Programmable page aligned address of the buffer
Head pointer
Programmable to initially setup ring
Hardware-maintained dword offset in the ring buffer. Pointer wraps.
DMA pointer
Hardware-maintained DMA request double-qword offset. Pointer wraps.
Tail pointer
Programmable double-dword offset in the ring buffer.
10.4.4.2
Ring Buffer Initialization
Before initializing a RB, the software first must allocate the desired number of 4-KB pages for use as buffer space. Then
the RINGBUF registers associated with the RB are programmed. Once the Ring Buffer Valid bit is set, the RB will be
considered for instruction arbitration, and the head and tail offsets will either indicate an empty RB (i.e., head == tail) or
will define some number of instructions to be executed.
10.4.4.3
Ring Buffer Use
Software can write new instructions into the “free space” of the RB, starting at the tail offset and up to (but not
including) the qword prior to the qword indicated by the head offset. (Recall that software must leave at least one qword
empty in the RB at all times). Note that this “free space” may wrap from the end of the RB back to the start.
Software must use some mechanism to track instruction execution progress in order to determine the “free space” in the
RB. This can be:
• A direct read of the Head Pointer Register
• The automatic reporting of the Head Pointer Register
• The explicit reporting of the Head Pointer Register via the GFXCMDPARSER_REPORT_HEAD instruction
• Some other “implicit” means by which software can determine how far the IP has progressed in retiring instructions
from an RB. This could include the use of “Store DWORD” instructions to write sequencing data to system
memory.
Once the instructions have been written (and padded out to a qword, if necessary), software can write the tail pointer
register to submit the new instructions for execution.
172
®
R
10.4.5
Batch Buffers
The GC provides for the execution of instruction sequences external to RBs. These sequences are called “batch buffers”
and are initiated through the use of GFXCMDPARSER_BATCH_BUFFER instructions that specify the starting address
and the length of the batch buffers. The arbitration rules used by the IP when executing batch buffers differ from those
employed when executing RBs, and are described later in this chapter. When a batch buffer instruction is executed out of
an RB, the initiated batch buffer sequence is such that the GC reads the instructions sequentially (via DMA) from the
batch buffer.
What happens when the end of the batch buffer is reached depends on the final instruction in the buffer. If the final
instruction is a GFXCMDPARSER_BATCH_BUFFER instruction, another batch buffer sequence is initiated. This
process, called “chaining,” continues until a batch buffer terminates with an instruction other than
GFXCMDPARSER_BATCH_BUFFER, at which point execution will resume in the RB at the instruction following the
initial GFXCMDPARSER_BATCH_BUFFER.
Instruction
From Ring Buffer
Instruction
Buffer
Chaining
Batch Buff Instr
Instruction
Instruction
Batch Buff Instr
Buffer
Chaining
Instruction
Instruction
Instruction
Return to Ring Buf
btch_buf.vsd
Figure 31. Batch buffer sequence
10.4.6
Instruction Arbitration
The Instruction Parser supports up to four sources of pending instructions: two Ring Buffers and two Batch Buffer
sequences (one batch buffer per ring buffer). The IP employs a set of rules to arbitrate among these instruction stream
sources. This section describes these rules and discusses the reasoning behind them.
10.4.6.1
Arbitration Rationale
The “Low-Priority” Ring Buffer (LPRB) is considered the primary mechanism utilized by drivers to pass instructions to
the GC. However, the insertion of instruction sequences into the LPRB must be a “synchronous” operation. That is, the
software must guarantee mutually exclusive access to the LPRB among contending sources (drivers). This ensures that
one driver does not corrupt another driver's partially completed instruction stream.
There are two general sources of requirements for asynchronous instruction generation and execution: interrupt handlers
and contending drivers. An interrupt handler may be invoked when a driver is in the process of inserting instructions into
the LPRB. To permit the interrupt handler to generate an instruction stream, the Interrupt Ring Buffer (IRB) is provided.
173
®
R
The IRB instruction stream is considered higher priority than the LPRB stream. This priority is manifested by how the
IRB is treated in the arbitration rules.
Note, however, that the software retains some control over this arbitration process. The
GFXCMDPARSER_ARB_ON_OFF instruction can be used from the LP instruction stream (RB or batch buffer) to
temporarily disable the IRB from arbitration. This can be used to define uninterruptible “critical sections” in the LP
stream (e.g., where a GC state needs to be protected from IRB instruction execution).
Another requirement for asynchronous instruction generation arises from competing (and asynchronous) drivers (e.g.,
“user-mode” driver libraries). In this case, it is desirable to allow these entities to construct instruction sequences in an
asynchronous fashion, via batch buffers. Synchronization then is only required to “dispatch” the batch buffers via
GFXCMDPARSER_BATCH_BUFFER instructions inserted into the LPRB. Batch buffers also can be initiated from the
IRBthough there the batch buffer is intended more for performance reasonswhere pre-generated (or “canned”)
instruction streams can be dispatched without having to copy them into the IRB.
LPRB batch buffers are considered uninterruptible for a number of reasons: IP complexity, GC context management, etc.
To provide gross limits to IRB latency, batch buffers are interruptible at “chain points” (i.e., between the
GFXCMDPARSER_BATCH_BUFFER that ends a batch and the start of the new batch). Using this mechanism,
software can segment batched instruction sequences into chains of smaller batches. Batch buffers initiated from the IRB
cannot, by definition, be interrupted. So, their chaining is of limited use.
10.4.6.2
Wait Instructions
The GFXCMDPARSER_WAIT_EVENT instruction is provided in order to allow instruction streams to be held pending
an asynchronous event. When executed directly from an RB, the IP will treat the RB as if it were empty, until the
specific event occurs. This will temporarily remove that RB from arbitration. Therefore, a wait instruction placed in the
IRB can allow LPRB activity to start/resume, even though there may (likely) be IRB instructions still pending.
If the instruction is executed from a batch buffer, the IP will simply wait for the event to occur, without performing
rearbitration. As this basically halts the IP for a length of time, the use of wait instructions in batch buffers (and the
impact on latency/performance) should be considered carefully by software developers.
10.4.6.3
Instruction Arbitration Points
The IP performs arbitration for instruction execution at the following points:
•
Continuously when idle (i.e., no pending instructions)
•
Between instructions in the LPRB
•
After GFXCMDPARSER_BATCH_BUFFER instructions when executed from the LPRB or at the end of an
LP batch buffer
•
Upon execution of a wait instruction in the IRB (if a wait is required)
The software must consider the consequences of the IP redirecting instruction execution at these arbitration points. That
is, the software must coordinate the use and control of the instruction stream sources in such a manner that the GC
operations proceed in the intended order and with the intended GC state. For example, software must prevent the case
where instructions placed in the IRB interrupt the LP instruction stream and invalidate the GC state required by the
pending LP stream.
174
®
R
10.4.6.4
Instruction Arbitration Rules
At an arbitration point, the IP will consider the current state of instruction execution (i.e., Low Priority vs. Interrupt,
Ring vs. Batch) along with the current state of the RBs and possibly pending “wait events.” The IP then will determine
how to proceed with execution, given this status information and the following instruction stream priorities (highest
priority to lowest):
•
Next instruction in batch buffer (regardless of the initiating ring buffer)
•
Next instruction pending in IRB (assuming that IRB arbitration is enabled)
•
Initiation of LP batch buffer (including resumption of interrupted LP batch chain)
•
Next instruction pending in LPRB
10.4.6.5
Batch Buffer Protected Mode
To ensure that the graphics controller does not corrupt system memory or graphics memory as the result of invalid
instructions from a batch buffer sequence, the batch buffer instruction has a flag that can be set to indicate that it is from
a non-trusted source.
When the IP processes a non-trusted batch buffer from one of the ring buffers, it does not allow any immediate store
dword instruction, because this instruction causes writes to system memory, not gathered through the GTT. The
protection mode (Protected or Unprotected) is set in the batch buffer instruction that is in the ring buffer. The protection
mode set persists throughout the batch buffer sequence, including batch buffers that are chained. Thus, a chained batch
buffer cannot re-enable writes to system memory.
If the IP detects an instruction that is disallowed in protected mode, it stores the header of the instruction, the origin of
the instruction, and an error code. In addition, such an event can generate an interrupt or a hardware write to system
memory, if enabled and unmasked. At this point the IP, can only be reactivated by a reset.
175
®
10.5 Instruction Format
GC instructions are defined with various formats. The first dword of all instructions is called the “header” dword. The
header contains the only field common to all instructions: the “client” field that determines the major GC unit that will
process the instruction data. The Instruction Parser examines the client field of each instruction in order to condition the
further processing of the instruction and route the instruction data accordingly. Valid client values are:
• Instruction parser (00h)
• 2D processor (02h)
• 3D processor (03h)
GC instructions vary in length, though they always are multiples of dwords. The length of an instruction is: (1) implied
by the client/opcode, (2) fixed by the client/opcode, yet included in a header field (so that the IP explicitly knows how
much data to copy/process) or (3) variable, with a field in the header indicating the total length of the instruction.
Note that GC instruction sequences require qword alignment and padding to qword length in order to be placed in ring
and batch buffers.
The following subsections provide a brief overview of the GC instructions, by client type. Figure 32 is a diagram of the
formats of the header dwords for all GC instructions. Table 9 is a list of instruction mnemonics, by client type.
10.5.1
Instruction Parser (IP) instructions are basically those instructions that do not require processing by the 2D or 3D
rendering/mapping engines. The functions performed by these instructions include:
• Control of the instruction stream (e.g., batch buffer commands, breakpoints, ARB On/Off, etc.)
• Hardware synchronization (e.g., flush, wait-for-event)
• Software synchronization (e.g., store DWORD, report head)
• Graphics buffer definition (e.g., display buffer, overlay buffer, 3D destination and Z buffer)
• Miscellaneous functions
10.5.2
2D Instructions
The 2D instructions include various flavors of Blt operations, along with instructions for setting up a blt engine state
without actually performing a Blt. Most instructions are of fixed length, though a few instructions include a variable
amount of “inline” data at the end of the instruction. Refer to the 2D Instructions chapter for a description of these
instructions.
10.5.3
3D Instructions
The 3D instructions are used to program the 3D pipeline state and perform 3D and stretch blt operations. All 3D state
instructions are of fixed length, while the rendering instructions are all variable length. Refer to the Rendering Engine
Instructions chapter for a description of the 3D instructions.
Figure 32. Instruction format for first dword
176
R
®
R
Bits
TYPE
Parser
31:29
000
28:24
23
001
2D
010
21:0
Opcode
Identification no./dword count
Instruction-dependent data
5:0 – Dword count
5:0 – Dword count
5:0 – Dword count
00h – NOP
0Xh – Single dword packets
1Xh – Two dword packets
2Xh – Store dword packets
3Xh – Ring/batch buffer packets
Reserved
22
Opcode
4:0 – Dword count
3DState24
011
Opcode – 00000 - 01111
23:0 – 24 state and mask bits
3DState24NP
011
Opcode – 10000 - 11000
23:0 – 24 non-pipelined state and mask bits
Reserved
011
Opcode – 11001 - 11011
3DState16
011
Opcode – 11100
23:19
Sub Opcode
00h - 7Fh
3DState16NP
011
Opcode – 11100
23:19
Sub Opcode
80h - FFh
3DStateMW
011
Opcode – 11101
(Multiple
dword)
3DStateMWNP
23:16
18:16 Texture
Map.
15:0 – 16
State and
mask bits
18:16 Scissor
Rect. No.
15:0 – 16
State and
mask bits
15:0 – Dword count
Sub Opcode
00h - 7Fh
011
Opcode – 11101
(Multiple
dword)
23:16
Sub Opcode
80h - FFh
3DBlock
011
Opcode – 11110
23:16
Sub Opcode
3DPrim
011
Opcode – 11111
23:16
Sub Opcode
Reserved
1XX
Notes:
SrcCopyImmBlt does not follow the 2D format.
The qualifier “NP” indicates that the state variable is non-pipelined and the render pipe is flushed before such a state
variable is updated. All the other state variables are pipelined (default).
Table 9. Graphics Controller Instructions
177
®
Client
00h - Command Parser
R
Instruction
GFXCMDPARSER_NOP_IDENTIFICATION
GFXCMDPARSER_BREAKPOINT_INTERRUPT
GFXCMDPARSER_USER_INTERRUPT
GFXCMDPARSER_WAIT_FOR_EVENT
GFXCMDPARSER_FLUSH
GFXCMDPARSER_CONTEXT_SEL
GFXCMDPARSER_DEST_BUFFER_INFO
GFXCMDPARSER_FRONT_BUFFER_INFO
GFXCMDPARSER_Z_BUFFER_INFO
GFXCMDPARSER_REPORT_HEAD
GFXCMDPARSER_ARB_ON_OFF
GFXCMDPARSER_OVERLAY_FLIP
GFXCMDPARSER_LOAD_SCAN_LINES_INCL
GFXCMDPARSER_LOAD_SCAN_LINES_EXCL
GFXCMDPARSER_STORE_DWORD_IMM
GFXCMDPARSER_STORE_DWORD_INDEX
GFXCMDPARSER_BATCH_BUFFER
02h - 2D Processor
SETUP_BLT
SETUP_MONO_PATTERN_SL_BLT
PIXEL_BLT
SCANLINE_BLT
TEXT_BLT
TEXT_Immediate_BLT
COLOR_BLT
PAT_BLT
MONO_PAT_BLT
SRC_COPY_BLT
SRC_COPY_Immediate_BLT
MONO_SRC_COPY_BLT
MONO_SRC_COPY_Immediate_BLT
FULL_BLT
FULL_MONO_SRC_BLT
FULL_MONO_PATTERN_BLT
FULL_MONO_PATTERN_MONO_SRC_BLT
03h - Rendering Processor
GFXPRIMITIVE
GFXRENDERSTATE_VERTEX_FORMAT
178
®
R
Table 9. Graphics Controller Instructions
Client
Instruction
GFXRENDERSTATE_MAP_TEXELS
GFXRENDERSTATE_MAP_COORD_SETS
GFXRENDERSTATE_MAP_INFO
GFXRENDERSTATE_MAP_FILTER
GFXRENDERSTATE_MAP_LOD_LIMITS
GFXRENDERSTATE_MAP_LOD_CONTROL
GFXRENDERSTATE_MAP_PALETTE_LOAD
GFXRENDERSTATE_MAP_COLOR_BLEND_STAGES
GFXRENDERSTATE_MAP_ALPHA_BLEND_STAGES
GFXRENDERSTATE_COLOR_FACTOR
GFXRENDERSTATE_SCR_DST_BLEND_MONO
GFXRENDERSTATE_Z_BIAS_ALPHA_FUNC_REF
GFXRENDERSTATE_LINE_WIDTH_CULL_SHADE_MODE
GFXRENDERSTATE_BOOLEAN_ENA_1
GFXRENDERSTATE_BOOLEAN_ENA_2
GFXRENDERSTATE_FOG_COLOR
GFXRENDERSTATE_DRAWING_RECTANGLE_INFO
GFXRENDERSTATE_SCISSOR_ENABLE
GFXRENDERSTATE_SCISSOR_RECTANGLE_INFO
GFXRENDERSTATE_ANTI_ALIASING
GFXRENDERSTATE_PROVOKING_VTX_PIXELIZATION_RULE
GFXRENDERSTATE_DEST_BUFFER_VARIABLES
179
®
R
180
®
R
11. Instruction Parser Instructions
11.1 Introduction
The Graphics Controller (GC) contains an extensive set of instruction for controlling 2D and 3D operations. This section
describes the programmer’s interface to these instructions, which can be categorized as follows:
• 3D instructions. The 3D pipeline states and processing functions are controlled by a set of 3D instructions. (See the
3D instruction section for detailed descriptions of the instructions.)
• 2D instructions. The 2D instructions are used to invoke BLT operations. (See the 2D register and instruction section
for detailed descriptions of the instructions).
• Instruction parser instructions. These instructions control various GC interface units (e.g., local memory interface,
display interface), setting breakpoints, etc.
11.2 Instruction Descriptions
11.2.1
GFXCMDPARSER_NOP_IDENTIFICATION
This instruction effectively provides a “no-operation” instruction that can be used to pad the instruction stream (e.g., in
order to pad out a batch buffer to a quadword boundary). However, there is one operation that this instruction can
perform. If the Enable bit is set, the command parser will write the Identification Number field contents into the NOP
Identification Register. This provides a general-purpose instruction stream tagging (“breadcrumb”) mechanism.
One possible example use would be for software to use a NOP_IDENTIFICATION instruction to tag a subsequent
breakpoint interrupt event. The GFXCMDPARSER_NOP_IDENTIFICATION instruction format is as follows:
DWord
Bit
0
31:29
Client: 000 – Instruction Parser
28:23
Opcode: 00h
22
Description
Enable:
1 = Write the identification no.
0 = Don’t write the identification no.
21:6
Identification no.
5:0
“Reserved MBZ”
181
®
11.2.2
R
GFXCMDPARSER_BREAKPOINT_INTERRUPT
This instruction will generate a breakpoint interrupt and cause the parser to stop until the interrupt is cleared, by writing
the Interrupt Identity Register. If the interrupting event is masked through the IMR, the parser will just continue parsing.
However, if the event is unmasked in the IMR and the interrupt is not enabled through the IER, the parser will halt until
the IIR is cleared. In this case, the software has to poll the IIR to check for this condition and clear the IIR (because an
interrupt will not occur).
DWord
Bit
0
31:29
28:23
Opcode: 01h
22:0
“Reserved MBZ”
11.2.3
Description
GFXCMDPARSER_USER_INTERRUPT
This instruction will generate a user-defined interrupt if the interrupt is enabled and not masked. The parser will continue
parsing after processing this instruction. If a user interrupt is currently outstanding (not yet cleared in the IIR), this
packet has no effect.
DWord
Bit
0
31:29
28:23
Opcode: 02h
22:0
“Reserved MBZ”
182
Description
®
R
11.2.4
GFXCMDPARSER_WAIT_FOR_EVENT
This instruction can be used to pause instruction stream processing until a specific event occurs. Only one event can be
specified; the specification of multiple events is UNDEFINED. The effect of the wait operation depends on the source
of the instruction. If the instruction is executed from a batch buffer, the command parser will halt (and suspend
instruction arbitration) until the event occurs. If it is executed from a ring buffer, further processing of that ring will be
suspended, although instruction arbitration (from other rings) will continue.
DWord
Bit
Description
0
31:29
28:23
Opcode: 03h
22:4
Reserved: 0000h
3
VBLANK: If this bit is set, and this instruction is executed out of a batch buffer, the parser
halts when it parses this instruction, until the beginning of the next display vertical blank. If it
is executed out of a ring buffer, the parser sets a flag that eliminates that ring from the
arbitration until the flag is cleared. This flag is cleared by the appropriate edge detection of
the Display Vertical Blank signal assertion.
2
DISPLAY FLIP PENDING: If this bit is set and this instruction is executed out of a batch
buffer, the parser halts when it parses this instruction, until the flip event. (The new front
buffer address has now been loaded into the active front buffer registers). If the instruction is
executed out of a ring buffer, the parser sets a flag that eliminates that ring from the
arbitration until the flip event. If there is no flip pending, the parser just proceeds.
Mechanism: The execution of a front buffer packet by the parser sets a flag
(FlipPendingFlag) that is cleared by the flip event. If the FlipPendingFlag is set and this
instruction shows up with this bit set, then the parser will wait for flip event, just as described
for the bit above. If FlipPendingFlag is not set and this instructions shows up with this bit set,
the instruction has no effect and the parser moves on. If the execution of this instruction
happens to coincide with the flip event, the parser behaves as if FlipPendingFlag is not set.
The flip event can be a function of hsync or vsync, as selected by the front buffer packet.
1
SCAN LINES: This instruction with this bit set should be sent after the LOAD_SCAN_LINES
instruction. If executed out of a batch buffer, this instruction will cause the parser to halt and
wait, if the scan_line_window indicator is asserted. If the scan line window indicator is deasserted, the parser just moves on. If it is executed out of a ring buffer, the parser sets a flag
if the scan line window indicator is asserted. This flag eliminates that ring from the arbitration
until it is cleared. This flag is cleared by the trailing edge of the scan_line_window indicator.
If the scan line window indicator is deasserted, the flag is not set and the parser moves on.
For more information look at the load_scan_line packet.
0
Undefined
183
®
11.2.5
R
GFXCMDPARSER_FLUSH
This instruction will flush all drawing engines and the frame buffer cache (aka local cache). In addition, it will
conditionally invalidate the map cache. After this instruction is completed and followed by a store dword, CPU access to
graphics memory will be coherent. In designs that implement an AGP connection, this instruction can cause an AGP
flush to be completed for the main memory control hub.
DWord
Bit
0
31:29
28:23
Opcode: 04h
22:2
Reserved: 0000h
184
Description
1
AGP FLUSH ENABLE: When this bit is set, the parser issues an AGP flush command and
waits for an acknowledge, before proceeding to the next command. (Not Implemented in the
Intel 82810 Chipset.)
0
INVALIDATE_MAP_CACHE: When this bit is set the parser will wait until the rendering
pipeline is done, after which it will invalidate the mapping engine cache.
®
R
11.2.6
GFXCMDPARSER_CONTEXT_SEL
The GFXCMDPARSER_CONTEXT_SEL instruction is used to pass pointers to the input interface in order to load or
use one of two sets of state variables. The pipeline supports the use of one set while the second set is being loaded. The
driver has to ensure that the pipeline is flushed before it changes the USE address. The LOAD address does not have the
same restriction. The format is as follows:
DWord
Bits
0
31:29
28:23
Opcode: 05h
22:18
Reserved: 00h
17
Description
Load Addr Enable:
1 = Update load address with the value in the Load Addr field.
0 = Ignore load address.
16
Use Addr Enable:
1 = Update use address with the value in the Use Addr field.
0 = Ignore Use Addr
15:9
8
Reserved for Load Addr: MBZ
Load Addr: Address of the SV set to be loaded
1 = State variable set to 1.
7:1
0
Reserved for Use Addr: MBZ
Use Addr: Address of the SV set to be used
185
®
11.2.7
R
The GFXCMDPARSER_DEST_BUFFER_INFO instruction is used to specify the information about the destination
buffer. This is an “immediate” instruction, so the software must guarantee that the rendering engine and the local
cache are flushed before modifying the destination buffer information. The format is as follows:
Dword
Bit
0
31:29
28:23
Opcode: 15h
22:6
Reserved: 0000h
5:0
DWORD_LENGTH: 00h
1
31:26
Reserved (additional address space)
25:12
Base Address: The base address of the rendered scene in linear space. The memory
interface unit uses this address in conjunction with the memory fence table registers, in
order to determine the virtual (tiled) address of the destination buffer. This surface address
(linear) must be at least 4-KB aligned. In addition, it must be 4-times-pitch aligned.
11:3
Reserved: 00h
2
Reserved MBZ
1:0
186
Description
Pitch: This is the pitch of the back buffer. This is used by the TLB in computing the absolute
address of the color requests.
00 – 512 bytes
01 – 1Kbytes
10 – 2Kbytes
11 – 4Kbytes
®
R
11.2.8
GFXCMDPARSER_FRONT_BUFFER_INFO
The GFXCMDPARSER_FRONT_BUFFER_INFO instruction is used to initialize the base address of the scene to be
displayed by the Display Engine (DE) (aka flip). There are two flavors of this instruction: In one, the instruction parser
sends the base address to the DE, where its update is synchronized to the display syncs (sync flip). In the second, the DE
update is on the following hsync (async flip).
In the case of a double-buffer swap operation requiring a flip between the display and render surface base addresses, in
addition to the FRONT_BUFFER_INFO packet, a DEST_BUFFER_INFO instruction must also be specified. It should
be noted that no special hardware is provided to synchronize these instructions.
A bit of the Interrupt Status Register represents the status of the flip instruction. This flag is set when the instruction
parser processes the GFXCMDPARSER_FRONT_BUFFER_INFO instruction. For the sync flip, the flag is cleared
when the vertical sync occurs. For the async flip the flag is cleared after the hardware determines that all the information
for the new buffer has been acquired. This is estimated to be 32 scan lines.
Setting and clearing this flag generates a system memory write to the location stored in the Hardware Status Vector
Address Register, if unmasked in the Hardware Status Mask Register. Clearing this flag will generate an external
interrupt, if unmasked in the Interrupt Mask Register and enabled in the Interrupt Enable Register.
The flush instruction should be issued prior to the flip instruction in order to ensure that hardware pipeline and cache
structures are flushed and that the rendered scene to be displayed next is in memory. Before parsing continues, the flush
instruction waits for the blitter and the render/map pipeline to be Not Busy and for the local cache to be coherent with
memory. The format of the GFXCMDPARSER_FRONT_BUFFER_INFO instruction is as follows:
DWord
Bits
0
31:29
28:23
Instruction Target: 14h
22:20
Reserved MBZ
19:8
Front Buffer Pitch: Will be loaded similarly to offset registers CR41(3:0):: CR13(7:0). Will
not be loaded in case of an async flip.
7
Reserved MBZ
6
Flip type: 0: synch flip, 1: async flip
5:0
1
Description
Dword Length: 00h
31:26
“Reserved MBZ”
25:3
Front Buffer Base Address: Virtual memory address bits 25:3 (max 64MB). The default
value is 0 (unsigned int).
2:0
“Reserved MBZ”
187
®
11.2.9
R
GFXCMDPARSER_Z_BUFFER_INFO
This instruction is used to specify the base address and pitch of the Z buffer surface used by the 3D rendering engine.
This is an “immediate” command, so the software must guarantee that the rendering engine and the local cache have
been flushed before modifying the z buffer information. The format of the GFXCMDPARSER_Z_BUFFER_INFO
instruction is as follows:
DWord
Bit
0
31:29
Client: 000 – Rendering Processor
28:23
Opcode: 16h
22:6
Reserved MBZ:.
5:0
Dword Length: 00h
1
188
Description
31:26
Reserved (Additional Address Space)
25:12
Base Address: The base address of the Z buffer in linear space. The Memory Interface
unit uses this address in conjunction with the Memory Fence Table Registers, in order to
determine the virtual (tiled) address of the buffer. This surface address (linear) must be at
least 4-KB aligned. in addition it must be 4-times-pitch aligned.
11:2
Reserved: 00h
1:0
Pitch: This is the pitch of the Z buffer. This is used by the TLB in computing the absolute
address of Z requests.
00 – 512 bytes
01 – 1 Kbyte
10 – 2 Kbytes
11 – 4 Kbytes
®
R
11.2.10
GFXCMDPARSER_REPORT_HEAD
This instruction causes the active ring buffer head pointer to be written to a cacheable (snooped) system memory
location. The location written is relative to the address programmed in the Hardware Status Page Address Register, and it
depends on which ring buffer is active. (Refer to the description of the HSW_PGA register.) The format is as follows:
DWord
Bit
0
31:29
28:23
Opcode: 07h
22:0
“Reserved MBZ”
11.2.11
Description
GFXCMDPARSER_ARB_ON_OFF
The GFXCMDPARSER_ARB_ON_OFF instruction is used to pass pointers to the input interface in order to turn on/off
all rings except the ring from which this instruction is executed. It can be used from a batch buffer. This instruction can
be used to prevent other ring buffers from interrupting an instruction sequence. The format is as follows:
DWord
Bits
0
31:29
28:23
Opcode: 08h
22:1
Reserved: 00h
0
Description
Arbitration ON/OFF:
1 = ON
0 = OFF
189
®
11.2.12
GFXCMDPARSER_OVERLAY_FLIP
DWord
Bits
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 00h
1
31:29
28:03
2:0
11.2.13
R
Description
“Reserved MBZ”
Register Update Address: This address is used by the overlay at the next VBLANK event,
in order to start requesting data from memory.
“Reserved MBZ”
GFXCMDPARSER_LOAD_SCAN_LINES_INCL
This instruction is used to initialize the scan line window registers in the display engine. If the display refresh is within
this window, the display engine asserts a signal that is used by the instruction parser to process the WAIT_FOR_EVENT
instruction. This instruction overrides any previous EXCL instruction. The format is as follows:
DWord
Bits
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 00h
1
190
Description
31:16
Start Scan Line Number (Line 0 is the first line of the display frame.)
15:0
End Scan Line Number
®
R
11.2.14
GFXCMDPARSER_LOAD_SCAN_LINES_EXCL
This instruction is used to initialize the scan line window registers in the display engine. If the display refresh is outside
this window, the display engine asserts a signal that is used by the instruction parser in order to process the
WAIT_FOR_EVENT instruction. This instruction overrides any previous INCL instruction. The format is as follows:
Word
Bits
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 00h
1
11.2.15
Description
31:16
Start Scan Line Number (Line 0 is the first line of the display frame.)
15:0
End Scan Line Number (Line 0 is the first line of the display frame.)
GFXCMDPARSER_STORE_DWORD_IMM
This instruction immediately causes a system write of the data word in the packet, to the address that also is in the
instruction packet. Note that all store dwords will invalidate the host-graphics pre-fetch cache. The format of the
GFXCMDPARSER_STORE_DWORD_IMM instruction is as follows:
DWord
Bit
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 01h
31:2
Address: DWORD aligned. The hardware only uses bits 31-2.
1:0
Reserved: MBZ
31:0
Data Word
1
2
Description
191
®
11.2.16
R
GFXCMDPARSER_STORE_DWORD_INDEX
This instruction immediately causes a system write of the data word in the packet, to the address provided in the
Hardware Status Page Address Register at the offset specified. Note that all store dwords will invalidate the hostgraphics pre-fetch cache.
DWord
Bit
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 01h
1
2
192
Description
31:12
Reserved: 0000h
11:2
Dword Offset into a page: A (11:2)
1:0
Reserved: 00
31:0
Data Word
®
R
11.2.17
GFXCMDPARSER_BATCH_BUFFER
The GFXCMDPARSER_BATCH_BUFFER instruction is used to pass pointers to the input interface in order to parse an
instruction buffer. The address on the instruction buffers is in graphics memory, and it should translate to a physical
address in main or local memory.
The batch buffer instruction packet implements a protection ID. The protection ID is recorded as the protection state by
the parser when it starts a batch buffer instruction executed from one of the rings. Chained batch buffer instructions
cannot change the protection state of the parser. If a batch buffer is pushed into the stack at a chain point, this state has to
be stored as well. The protection state modifies the parser behavior as follows:
In the Unprotected Mode, the parser reports an error if it parses a store dword immediate instruction. It is assumed that
an unprotected batch buffer has not been blessed by the driver and can have bogus store dword immediate instructions
that overwrite protected areas in cacheable memory. In the protected state, the parser will allow all instructions to be
parsed. It is assumed that a batch buffer identified as protected has been blessed by the driver, in order to ensure that
store dword immediate instructions do not corrupt the operating system.
DWord
Bits
0
31:29
28:23
22:6
Reserved: 0000h
5:0
Dword Length: 01h
31:3
Buffer Start Address: Must be quadword aligned in memory.
2:1
Reserved: 00h
1
0
2
Description
Protection id: 1 – Unprotected, 0 – Protected
31:3
Buffer End Address: Must be quadword aligned in main memory.
2:0
Reserved: MBZ
Helpful hint: Use GFXCMDPARSER_NOP_IDENTIFICATION to pad the buffer to quadword length.
193
®
R
194
R
®
12. 2D Instructions
12.1 Introduction
This chapter contains the 2D graphics controller instructions. For each instruction the format specifies the functionality
of field. When an instruction does not require a field, it is ignored. All registers can only be written through instructions.
No program I/O writing of the BLT registers is allowed.
12.2 BLTs to and from Cacheable Memory
The blitter can be used to transfer data from cacheable memory to graphics memory and vice-versa, by using the blitter
command packets. The source or destination operands in these packets can be steered towards cacheable memory.
Patterns may be used with the source. The driver is required to flush the drawing pipelines before and after each copy
command targeting cacheable memory. In addition, the driver is required to turn arbitration off if this command is used
from the low-priority ring buffer. An example sequence from the low-priority ring buffer is:
ARB_OFF, FLUSH, SCB, .........,FLUSH, SCB, FLUSH, ARB_ON
where all source copy blits (SCBs) use cacheable memory.
Either the source or destination surface can be in cacheable memory. It is not allowed to have both source and
destination surfaces in cacheable memory as part of the same blit operation. In either case, the surface address
programmed in this instruction must be in graphics address space. The GTT must be programmed to set up a scattergather translation from graphics memory pages to physical pages in cacheable memory. A surface that is being mapped
to cacheable space must not be tiled (hence not fenced).
Further restrictions are that the source data must be the same pixel width as the destination. The only function permitted
is color copies with a positive destination pitch and direction. The source operand can be either sign to allow mirroring in
the vertical direction. An immediate source operand is not allowed.
195
®
R
12.3 BLT Engine Instructions
The following instructions are directed to the BLT engine. The Instruction Target field is used as an opcode by the BLT
engine state machine to qualify which control bits are relevant for executing the instruction. The descriptions for each
dword and bit field are contained in the BLT Engine Instruction Definition section. Each dword in a packet has a
corresponding Instruction Definition description, where the details of the operation of the function are described.
NOTE: All reserved fields must be programmed to 0s.
12.3.1
SETUP_BLT
The setup instruction supplies common setup information, including clipping coordinates used exclusively with the
following three instructions:
PIXEL_BLT (PB)- 1 pixel write with the coordinate and solid pattern supplied for each pixel to be written. Neither
non-solid patterns nor source operands are allowed.
SCANLINE_BLT (SLB) - 1 scan line of color or mono pattern and destination are the only operands allowed.
TEXT_BLT (TB) - Linear monochrome source either through the instruction stream or from graphics memory (not
cacheable) and the destination are the only operands allowed. Source copy is the only supported operation. The
raster operation field must = CC hex.
Clipping addresses and coordinates are inclusive. (The BLT engine performs a trivial reject for all three of these BLT
instructions before performing any accesses. If any pixels are included within the clipping rectangle, then it performs
every access, but deasserts the byte enables for the pixels that are clipped.)
The source operand (TEXT_BLT only) never overlaps the destination. Therefore, the X and Y direction is always
positive (left to right and top to bottom). Only a positive destination pitch is allowed.
All fields above indicate which instructions require a valid state. These are the only instructions that require that the state
be saved between instructions. There are 4 dedicated registers that contain the states for these 3 instructions. All other
BLTs use a temporary version of these. The 4 double-word registers are DW1 (control), DW5 (foreground color), DW6
(background color), and DW7 (pattern address). The ClipRect registers do not need to be saved since the interrupt ring
buffer never uses them.
Notes:
The mono source transparency mode flag is maintained (although it is always set to one in current driver operation).
196
®
R
DWord
Bit
Description
0 = BR00
31:29
Client : 02h - 2D Processor
28:22
Instruction Target (Opcode) : 00h
21:05
Reserved
0
04:00
Dword Length : 06h
1 = BR01
31
Reserved
30
Reserved
29
Mono Source Transparency Mode: (1 = transparency enabled; 0 = use background) - TB
only
28
Reserved
27
Reserved
26
Dynamic Color Enable: (1 = use Color Depth field; 0 = default color depth) - All
25:24
Color Depth: All
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation: (all; must = CC Hex for TB)
15:00
Destination Pitch (positive): (13:00 are implemented in Intel 82810 Chipset.) (SLB & TB
only)
2 = BR02
31:00
ClipRect Y1 Address (Top): All (25:00 are implemented in Intel 82810 Chipset.)
3 = BR03
31:00
ClipRect Y2 Address (Bottom): All (25:00 are implemented in Intel 82810 Chipset.)
4 = BR04
31:16
ClipRect X2 Coordinate (Right): All (Intel 82810 Chipset implementation supports 27:16 =
12 bits)
15:00
ClipRect X1 Coordinate (Left): All (Intel 82810 Chipset implementation supports 11:00 =
12 bits)
5
31:24
Reserved
5 = BR05
23:00
Setup Background Color: All
6
31:24
Reserved
6 = BR06
23:00
Setup Foreground Color: (SLB & TB only)
7 = BR07
31:00
Pattern Address for Color Pattern: (25:06 are implemented in Intel 82810 Chipset.)
(SLB only)
®
®
®
®
®
®
197
®
12.3.2
R
SETUP_MONO_PATTERN_SL_BLT
This setup instruction supplies common setup information, including clipping coordinates used exclusively with the
following instruction:
SCANLINE_BLT (SLB) - 1 scan line of monochrome pattern and destination are the only operands allowed.
Clipping addresses and coordinates are inclusive. (The BLT engine performs a trivial reject for this BLT before
performing any accesses. If any pixels are included within the clipping rectangle, then it performs every access, but
deasserts the byte enables for the pixels that are clipped.) These are the only instructions the require that the state be
saved between instructions. There are 4 dedicated registers to contain the state for this instruction. The 3 double-word
registers are: DW1 (control), DW5 (background color), and DW6 (foreground color). The ClipRect registers do not need
to be saved since the interrupt ring buffer never uses them.
Only a positive destination pitch is allowed.
DWord
Bit
0 = BR00
31:29
28:22
21:05
Reserved
04:00
Dword Length : 07h
31
Solid Pattern Select: (1 = solid pattern; 0 = no solid pattern) - (SLB & Pixel only)
30:29
Reserved
28
Mono Pattern Transparency Mode: (1 = transparency enabled; 0 = use background)
27
Reserved
26
Dynamic color Enable: (1 = use Color Depth Field; 0 = default color depth)
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
Destination Pitch (positive): (13:00 are implemented in Intel 82810 Chipset.)
2 = BR02
31:00
ClipRect Y1 Address (Top): (25:00 are implemented in Intel 82810 Chipset.)
3 = BR03
31:00
ClipRect Y2 Address (Bottom): (25:00 are implemented in Intel 82810 Chipset.)
4 = BR04
31:16
ClipRect X2 coordinate (Right): (Intel 82810 Chipset implementation supports 27:16 = 12
bits)
15:00
ClipRect X1 Coordinate (Left): (Intel 82810 Chipset implementation supports 11:00 = 12
bits)
5
31:24
Reserved
5 = BR05
23:00
Setup Background Color: All
6
31:24
Reserved
6 = BR06
23:00
Setup Foreground Color: (SLB & TB only)
7 = BR20
31:00
DW0 (least significant) for a Monochrome Pattern:
8 = BR21
31:00
DW1 (most significant) for a Monochrome Pattern:
1 = BR01
198
Description
®
®
®
®
®
®
R
12.3.3
PIXEL_BLT
The destination X coordinate and destination Y address are compared with the ClipRect registers. If they are within the
ClipRect bounds, then the pixel supplied in the SETUP_BLT instruction is written with the raster operation to
(destination Y address + destination X coordinate * bytes per pixel).
DWord
Bit
0 = BR00
31:29
28:22
21:06
Destination X Coordinate: (Intel 82810 Chipset implementation supports 17:06 = 12 bits)
05
Reserved
0
04:00
Dword Length : 00h
1 = BR09
31:00
Destination Y Address: (address of the first pixel on a scan line) (25:00 are implemented in
®
Intel 82810 Chipset.)
0 = BR08
Description
®
199
®
12.3.4
R
SCANLINE_BLT
The destination Y address is compared with the ClipRect Y address registers. If the address is within the ClipRect
bounds, then the pattern pixels dependent on the destination X coordinates that fall within the ClipRect X bounds are
written using the raster operation to (destination Y address + destination X coordinate * bytes per pixel). The horizontal
alignment is relative to the destination from the lower bits of the destination address. The pattern vertical alignment
indicates which pattern scan line is used. (This is the least-significant three bits of the destination vertical coordinate.)
With color patterns, only 1 scan line should be read for this instruction.
Solid pattern should use the SETUP_MONO_PATTERN_SL_BLT instruction.
DWord
0 = BR00
1 = BR08
Bit
Description
31:29
28:22
21:08
Reserved
07:05
Pattern Vertical Alignment: (which scan line of the 8x8 pattern to start on)
04:00
Dword Length : 01h
31:16
Destination X2 Coordinate: (Ending - Right)
®
(Intel 82810 Chipset implementation supports 27:16 = 12 bits)
15:00
Destination X1 Coordinate: (Starting - Left) – X2 – X1 + 1 = width in pixels
®
2 = BR09
31:00
Destination Y Address: (address of the first pixel on a scan line)
®
(25:00 is implemented in Intel 82810 Chipset)
200
®
R
12.3.5
TEXT_BLT
All monochrome source scan lines and pixels that fall within the ClipRect Y addresses and X coordinates are written
(ignoring the raster operation) to (destination Y address + destination X coordinate * bytes per pixel). Source expansion
color registers are always in SETUP_BLT.
Notes:
All graphics controller BR03 fields (monochrome clipping parameters) are computed by the hardware while performing
clipping.
DWord
Bit
0 = BR00
31:29
28:22
21:17
Reserved
16
Bit (0) / Byte (1) packed: Byte packed is for the NT driver.
04:00
Dword Length : 04h
31:16
1 = BR08
Description
®
15:00
Destination X1 Coordinate: (Starting - Left) - X2 - X1 + 1 = width in pixels
®
2 = BR09
31:00
Destination Y1 Address: (address of the first pixel on the first scan line)
®
(25:00 are implemented in Intel 82810 Chipset.)
3 = BR10
31:00
Destination Y2 Address: (address of the first pixel on the last scan line)
®
4= BR11
5 = BR12
31:16
Reserved
15:00
Number of monochrome source quadwords - 1: (1 to 64k quadwords = 64 bits to 4
Mbits)
31:00
Source Address: (address of the first byte of the first pixel on the first scan line)
®
(25:00 are implemented in Intel 82810 Chipset
201
®
12.3.6
R
TEXT_Immediate_BLT
This instruction allows the driver to send data through the instruction stream, which eliminates the read latency when
reading a source from memory. This allows graphics primitives such as Text to execute much faster. If an operand is in
system-cacheable memory and is either small or only accessed once, it can be copied directly to the instruction stream
instead of to graphics-accessible memory.
The IMMEDIATE_BLT data MUST transfer an even number of DWs. The BLT engine will hang if it does not get an
even number of DWs.
Monochrome source data is sent through the instruction stream.
All monochrome source scan lines and pixels that fall within the ClipRect Y addresses and X coordinates are written
(ignoring the raster operation) to (destination Y address + destination X coordinate * bytes per pixel). Source expansion
color registers are always in the SETUP_BLT.
DWord
Bit
0 = BR00
31:29
28:22
21:17
Reserved
16
Bit (0) / Byte (1) packed: Byte packed is for the NT driver.
15:00
Dword Length : 02+ DWL = (Number of Immediate double words) h
31:16
1 = BR08
Description
®
(Intel 82810 Chipset implementation supports 27:16 = 12 bits.)
15:00
Destination X1 coordinate: (Starting - Left) - X2 - X1 + 1 = width in pixels
®
(Intel 82810 Chipset implementation supports 11:00 = 12 bits.)
2 = BR09
31:00
Destination Y1 Address: (address of the first pixel on the first scan line)
®
3 = BR10
31:00
Destination Y2 Address: (address of the first pixel on the last scan line)
®
4
31:00
Immediate Data DW 0:
5
31:00
6 thru
DWL+3
S
Immediate Data DWs 2 through DWORD_LENGTH (DWL):
202
®
R
12.3.7
COLOR_BLT
COLOR_BLT is the simplest BLT operation. It performs a color fill to the destination (with a possible ROP). The only
operand is the destination operand, which is written according to the raster operation. The solid pattern color is stored in
the pattern background register.
This instruction is optimized to run at the maximum memory write bandwidth.
The typical raster operation code = F0, which copies the pattern background register to the destination.
DWord
Bit
0 = BR00
31:29
28:22
21:05
Reserved
04:00
Dword Length : 03h
31
Solid pattern select: (1 = solid pattern). Must be 1.
30:27
Reserved
1 = BR13
2 = BR14
3 = BR09
Description
26
Dynamic Color Enable: (1 = use Color Depth Field; 0 = default color depth)
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = Reserved
23:16
Raster Operation:
15:00
31:16
Destination Height (in scan lines): (28:16 are implemented in Intel 82810 Chipset.)
15:00
Destination Width (in bytes): (12:00 are implemented in Intel 82810 Chipset.)
31:00
Destination Address: Address of the first byte to be written
®
®
®
®
4 = BR16
31:24
Reserved
23:00
Solid Pattern Color:
203
®
12.3.8
R
PAT_BLT
PAT_BLT is used when there is no source and the color pattern is not trivial (i.e., is not a solid color only).
The whole color pattern (8 × 8 pixels = 16, 32 or 64 DWs) is read at the beginning of the BLT and stored in the Texture
Cache. The pattern vertical alignment specifies the first scan line of the pattern that is used. The horizontal alignment is
relative to the destination, from the lower bits of the destination address.
The only memory accesses required for the remainder of the BLT are destination accesses, which are dependent on the
raster operation.
DWord
Bit
0 = BR00
31:29
28:22
21:08
Reserved
07:05
Pattern Vertical Alignment:
04:00
Dword Length : 03h
31:28
Reserved
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
3 = BR09
31:00
®
4 = BR15
31:00
Pattern Address: (25:06 are implemented in Intel 82810 Chipset.)
1 = BR13
2 = BR14
204
Description
®
®
®
®
®
R
12.3.9
MONO_PAT_BLT
MONO_PAT_BLT is used when there is no source and the monochrome pattern is not trivial (i.e., is not a solid color
only). The monochrome pattern is loaded from the instruction stream and the only memory accesses are for the
destination operand, which is dependent on the raster operation. The pattern vertical alignment indicates the byte at
which to start. The horizontal alignment is relative to the destination, from the lower bits of the destination address. The
monochrome pattern transparency mode indicates whether to use the pattern background color or deassert the write
enables when the bit in the pattern is 0. When the pattern bit is 1, then the pattern foreground color is used in the ROP
operation. The ROP value chosen should involve the mono pattern data in the ROP operation.
DWord
Bit
0 = BR00
31:29
28:22
21:08
Reserved
07:05
Pattern Vertical Alignment
04:00
Dword Length : 06h
31:29
Reserved
28
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
31:00
®
31:24
Reserved
23:00
Pattern Background Color:
31:24
Reserved
5 = BR17
23:00
Pattern Foreground Color:
6 = BR20
31:00
Pattern Data 0: (least-significant DW)
7 = BR21
31:00
Pattern Data 1: (most-significant DW)
1 = BR13
2 = BR14
3 = BR09
4 = BR16
Description
®
®
®
205
®
12.3.10
R
SRC_COPY_BLT
This BLT instruction performs a color source copy when the only operands involved are a color source and destination
of the same bit width.
The source and destination operands may overlap, which means that the X and Y directions can be either forwards or
backwards. The X direction field applies to both the destination and source operands. The source and destination pitches
can be either sign.
DWord
Bit
0 = BR00
31:29
28:22
21:05
Reserved
04:00
Dword Length : 04h
31
Reserved
30
X Direction (1 = written from right to left (decrementing = backwards); 0 = incrementing)
29:28
Reserved
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
Destination Pitch (signed): (13:00 are implemented in Intel 82810 Chipset.)
31:16
15:00
31:00
®
31:14
Reserved
4 = BR11
13:00
Source Pitch (quadword aligned and signed): (13:00 are implemented in Intel 82810
Chipset.)
5 = BR12
31:00
Source Address: (25:00 are implemented in Intel 82810 Chipset.)
1 = BR13
2 = BR14
3 = BR09
206
Description
®
®
®
®
®
®
R
12.3.11
SRC_COPY_IMMEDIATE_BLT
reading a source from memory. This allows graphics primitives such as Text to execute much faster. If an operand is in
system-cacheable memory and is either small or only accessed once, it can be copied directly to the instruction stream
instead of to graphics accessible memory.
The IMMEDIATE_BLT data MUST transfer an even number of DWs. The BLT engine will hang if it doesn't get an
even number of DWs.
This BLT instruction performs a color source copy where the only operands involved are a color source and destination
of the same bit width.
Immediate source data is quadword aligned, which means throwing away all bytes at the end of a scan line, until the next
naturally aligned double word.
Restriction: Immediate operand must be ≤ 16 QWs.
DWord
0 = BR00
1 = BR13
Bit
Description
31:29
28:22
21:16
Reserved
15:00
Dword Length : 02+ DWL = (Number of Immediate double words = maximum number of
quadwords for the GMCH is 16 = 32 doublewords) h
31
Reserved
30
29:28
Reserved
27
Source Select Mode: = 1. Must be 1.
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
3 = BR09
31:00
®
4
31:00
5
31:00
6 thru
DWL+3
S
2 = BR14
®
®
®
207
®
12.3.12
MONO_SRC_COPY_BLT
This BLT instruction performs a monochrome source copy in which the only operands are a monochrome source and
destination. The source and destination operands cannot overlap, which means that the X direction must always be
forward.
All non-text monochrome sources are word aligned. At the end of a scan line, the monochrome source bits until the next
word boundary must be ignored. The monochrome source data bit position field <2:0> indicates which bit position
within the first byte of the scan line should be used as the first source pixel.
The monochrome source transparency mode indicates whether to use the source background color or deassert the write
enables when the bit in the source is 0. When the source bit is 1, the source foreground color is used in the ROP
operation. The ROP value chosen should involve the mono source data in the ROP operation.
The destination pitch can either be positive or negative, in order to allow mirroring in the Y direction.
208
R
®
R
DWord
0 = BR00
1 = BR13
Bit
Description
31:29
28:22
21:20
Reserved
19:17
Monochrome source data bit position of the first pixel within a byte per scan line
16:05
Reserved
04:00
Dword Length : 06h
31:30
Reserved
29
Mono Source Transparency Mode: (1 = transparency enabled; 0 = use background)
28
Reserved
27
Source Select Mode: = 1
26
Dynamic color Enable: (1 = use Color Depth Field; 0 = default color depth)
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
3 = BR09
31:00
®
4
31:16
Reserved
4 = BR11
15:00
Number of Monochrome Source Quadwords - 1: (1 to 64k Quadwords = 64 bits to 4
Mbits)
5 = BR12
31:00
®
6
31:24
Reserved
6 = BR18
23:00
Source Background Color:
7
31:24
Reserved
7 = BR19
23:00
Source Foreground Color:
2 = BR14
®
®
®
209
®
12.3.13
R
MONO_SRC_COPY_IMMEDIATE_BLT
reading a source from memory. If an operand is in system-cacheable memory and is either small or only accessed once, it
can be copied directly to the instruction stream instead of to graphics accessible memory.
The IMMEDIATE_BLT data MUST transfer an even number of DWs. The BLT engine will hang if it doesn't get an
even number of DWs or the exact number of QWs required.
This BLT instruction performs a monochrome source copy where the only operands involved are a monochrome source
and destination. The source and destination operands cannot overlap, which means that the X direction must always be
forward.
Monochrome source data is sent through the instruction stream. BR11 = DW0 [15:0] - 4
within the first byte should be used as the first source pixel.
enables when the bit in the source is 0. When the source bit is 1, then the source foreground color is used in the ROP
The Destination pitch can either be positive or negative, in order to allow mirroring in the Y direction.
210
®
R
DWord
Bit
0 = BR00
31:29
28:22
21:20
Reserved
19:17
16
Reserved
0 = BR11
15:00
Dword Length : 04+ DWL = (Number of immediate double words) h
1 = BR13
31:30
Reserved
29
28
Reserved
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
31:00
®
31:24
Reserved
4 = BR18
23:00
5
31:24
Reserved
5 = BR19
23:00
6
31:00
7
31:00
8 thru
DWL+4
S
2 = BR14
3 = BR09
Description
®
®
®
211
®
12.3.14
R
FULL_BLT
The full BLT is the most comprehensive BLT instruction. It provides the ability to specify all 3 operands: destination,
source, and pattern. The source and pattern operands are the same bit width as the destination operand.
Cache. The pattern vertical alignment specifies which scan line of the pattern is used first. The destination address
specifies the horizontal alignment The only memory accesses required for the remainder of the BLT are the source and
destination accesses.
Both the source and destination pitches can be either sign. The pattern direction follows the destination operand.
DWord
Bit
0 = BR00
31:29
1 = BR13
Description
28:22
21:11
Reserved
10:08
Destination Transparency Mode:
07:05
04:00
Dword Length : 06h
31
Reserved
30
X Direction: (1 = written from right to left (decrementing = backwards); 0 = incrementing)
29:28
Reserved
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
31:00
®
31:14
Reserved
4 = BR11
13:00
Source Pitch (quadword aligned and signed): (13:00 are implemented in Intel 82810
Chipset.)
5 = BR12
31:00
2 = BR14
3 = BR09
®
®
®
®
®
31:24
Reserved
6 = BR18
23:00
Destination Transparency Color:
7 = BR15
31:00
212
®
®
R
12.3.15
FULL_MONO_SRC_BLT
source, and pattern. The source operand is monochromatic and the pattern operand is the same bit width as the
destination operand.
Cache. The pattern vertical alignment specifies which scan line of the pattern is used first. The destination address
specifies the horizontal alignment The only memory accesses required for the remainder of the BLT are the destination
and sometimes the monochrome source accesses, since the source is monochromatic.
enables when the bit in the source is 0. When the source bit is 1, then the source foreground color is used in the ROP
All non-text and non-immediate monochrome sources are word aligned. At the end of a scan line, the monochrome
source bits until the next word boundary must be ignored. The monochrome source data bit position field [2:0] indicates
which bit position within the first byte should be used as the first source pixel.
The destination pitch can be either sign, in order to allow the vertical mirroring of a monochrome source with a pattern
that is accessed in the same direction as the destination operand.
213
®
DWord
Bit
R
Description
31:29
28:22
21:20
Reserved
19:17
16:08
Reserved
07:05
04:00
Dword Length : 07h
31:30
Reserved
29
28
Reserved
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster operation:
15:00
31:16
15:00
31:00
31:16
Reserved
4 = BR11
15:00
Number of Monochrome Source Quadwords - 1: (1 to 64k Quadwords = 64 bits to 4
Mbits)
5 = BR12
31:00
®
31:24
Reserved
23:00
31:24
Reserved
7 = BR19
23:00
8 = BR15
31:00
0 = BR00
1 = BR13
2 = BR14
3 = BR09
6 = BR18
214
®
®
®
®
®
R
12.3.16
FULL_MONO_PATTERN_BLT
source, and pattern. The pattern operand is monochrome and the source operand is the same bit width as the destination
operand.
The monochrome pattern is loaded from the instruction stream. The pattern vertical alignment specifies which scan line
of the pattern is used first. The destination address specifies the horizontal alignment The only operands accessed from
memory are the source and destination operands.
The monochrome pattern transparency mode indicates whether to use the pattern background color or deassert the write
enables when the bit in the source is 0. When the source bit is 1, then the pattern foreground color is used in the ROP
operation. The ROP value chosen should involve the mono pattern data in the ROP operation.
Both the source and destination pitches can be either sign. The pattern direction follows the destination operand.
215
®
DWord
Bit
0 = BR00
31:29
28:22
21:11
Reserved
R
Description
10:08
07:05
04:00
Dword Length : 09h
31
Solid Pattern Select: (1 = solid pattern; 0 = no solid pattern)
30
X Direction: (1 = written from right to left (decrementing = backwards; 0 = incrementing)
29
Reserved
28
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
31:00
®
31:16
Reserved
4 = BR11
15:00
Source Pitch: (quadword aligned and signed)
®
5 = BR12
31:00
6 = BR18
31:00
Destination Transparency Color:
31:24
Reserved
23:00
31:24
Reserved
8 = BR17
23:00
9 = BR20
31:00
A =BR21
31:00
1 = BR13
2 = BR14
3 = BR09
7 = BR16
216
®
®
®
®
®
R
12.3.17
FULL_MONO_PATTERN_MONO_SRC_BLT
The full BLT provides the ability to specify all 3 operands: destination, source, and pattern. The pattern and source
operands are monochromatic.
The monochrome pattern is loaded from the instruction stream. The pattern vertical alignment specifies which scan line
of the pattern is used first. The destination address specifies the horizontal alignment The only memory accesses required
for the remainder of the BLT are the destination and monochrome source accesses.
enables when the bit in the source is 0. When the source bit is 1, the source foreground color is used in the ROP
operation. The ROP value chosen should involve the mono source and mono pattern.
within the first byte should be used as the first source pixel.
The monochrome pattern transparency mode indicates whether to use the pattern background color or deassert the write
enables when the bit in the source is 0. When the source bit is 1, then the pattern foreground color is used in the ROP
operation. The monochrome source transparency mode works identically to the pattern transparency mode.
The destination pitches can be either sign. The pattern direction follows the destination operand.
217
®
DWord
Bit
R
Description
31:29
28:22
21:20
Reserved
19:17
16:08
Reserved
07:05
04:00
Dword Length : 0Ah
31
Solid Pattern Select: (1 = solid pattern; 0 = no solid pattern)
30
Reserved
29
28
27
Reserved
26
25:24
Color Depth:
00 = 8-bit color
01 = 16-bit color
10 = 24-bit color
11 = reserved
23:16
Raster Operation:
15:00
31:16
15:00
31:00
®
31:16
Reserved
4 = BR11
15:00
Number of Monochrome Source Quadwords - 1:
(1 to 64k Quadwords = 64 bits to 4 Mbits)
5 = BR12
31:00
31:24
Reserved
23:00
31:24
Reserved
23:00
31:24
Reserved
23:00
31:24
Reserved
9 = BR17
23:00
A =BR20
31:00
B =BR21
31:00
0 = BR00
1 = BR13
2 = BR14
3 = BR09
6 = BR18
7 = BR19
8 = BR16
218
®
®
®
®
®
R
12.4 BLT Engine Instruction Definitions
This section describes the BLT engine instruction fields. These descriptions are in the format of register descriptions. For
debug purposes, Read Only addresses provide the BLT engine status.
12.4.1
BR00—BLT Opcode & Control
Default:
Attributes:
40000h
0000 0000
RO; dword accessible
BR00 is the last-executed instruction dword 0. Bits [22:5] are written by every DW0 of every instruction. Bits [31:30]
and [4:0] are status bits. Bits [28:27] are written from DW0 [15:14] of a Setup instruction, and Bit 29 is written with a 1
whenever a Setup instruction is written. Bit 29 is a decode of the Setup instruction opcode.
31
30
29
BLT
BSY
Clip
Inst
Setup
Mono
Pattern
15
14
Reserved
28
22
Instruction Target (Opcode)
13
12
11
Text
BLT
Scan
Line
BLT
Pixel
BLT
10
8
Destination
Transparency
Mode
7
Pattern
Vertical
Alignment
21
20
Reserved
5
19
17
Monochrome
Source Start
16
Bit (0) /
Byte
(1)
Packed
4
3
2
1
0
DST
RMW
Color
Source
Mono
Source
Color
Pattern
Mono
Pattern
219
®
R
Bit
Descriptions
31
BLT Engine Busy. This bit indicates whether the BLT engine is busy (1) or idle (0). This bit is replicated in
the SETUP BLT Opcode & Control register.
1 = Busy
0 = Idle
30
Clip Instruction. The current instruction performs clipping (1).
29
Setup Monochromatic Pattern. This bit is decoded from the Setup instruction opcode, in order to identify
whether a color (0) or monochrome (1) pattern is used with the SCANLINE_BLT instruction.
1 = Monochromatic
0 = Color
28:22
Instruction Target (Opcode). This is the content of the Instruction Target field from the last BLT
instruction. This field is used by the BLT engine state machine to identify the BLT instruction to execute.
The opcode specifies whether the source and pattern operands are color or monochromatic.
21:20
Reserved
19:17
Monochrome Source Start. This field indicates the starting monochrome pixel bit position within a byte
per scan line of the source operand. The monochrome source is word aligned, which means that at the end
of the scan line all bits should be discarded until the next word boundary.
16
Bit/Byte Packed. Byte packed is for the NT driver
0 = Bit
1 = Byte
15:14
Reserved
13
Text BLT. Current Opcode is Text BLT.
12
Scan Line BLT. Current Opcode is Scan Line BLT.
11
Pixel BLT. Current Opcode is Pixel BLT.
220
®
R
Bit
Descriptions
10:8
Destination Transparency Mode. These bits control whether or not the byte(s) at the destination
corresponding to a given pixel will be conditionally written and what those conditions are. This feature can
make it possible to perform various masking functions in order to selectively write or preserve graphics data
already at the destination.
All four sets of conditions that may be chosen as the controlling factor in performing color transparency
involve comparing with other colors the color that has been specified for use in the color expansion of any
monochrome source data. This background color is in the Source Expansion Background Color Register.
XX0 = No color transparency mode enabled. This causes normal operation with regard to writing data to
the destination.
001 = The color specified as the background color for use in the color expansion of monochrome source
data is compared with the color resulting from the bit-wise operation performed for each pixel. If
these two colors are not equal, then the byte(s) at the destination corresponding to the current pixel
are written with the result of the bit-wise operation. [Source color transparency]
data is compared with the color specified by the byte(s) at the destination corresponding to the
current pixel. If these two colors are not equal, then the byte(s) at the destination corresponding to
the current pixel are written with the result of the bit-wise operation.
data is compared with the color resulting from the bit-wise operation performed for each pixel. If
these two colors are equal, then the byte(s) at the destination corresponding to the current pixel are
written with the result of the bit-wise operation.
data is compared with the color specified by the byte(s) at the destination corresponding to the
current pixel. If these two colors are equal, then the byte(s) at the destination corresponding to the
current pixel are written with the result of the bit-wise operation. [Destination color transparency]
7:5
Pattern Vertical Alignment. Specifies the scan line’s worth (i.e., which one of the 8 horizontal rows) of the
8x8 pattern that will appear on the first scan line’s worth of the data written to the destination. Depending
upon the location of the destination, the upper-left-hand corner of the upper-left-hand tile of the pattern is
usually aligned with the upper-left-hand corner of the block of data written to the destination. The BLT
engine determines the horizontal alignment relative to the destination from the lower bits of the destination
address. However, the vertical alignment relative to the destination must be supplied through these bits.
4
Destination Read Modify Write. This bit is decoded from the last instruction’s opcode field and
Destination Transparency Mode, in order to identify whether a destination read is needed.
3
Color Source. This bit is decoded from the last instruction’s opcode field to identify whether a color (1)
source is used.
2
Monochrome Source. This bit is decoded from the last instruction’s opcode field to identify whether a
monochrome (1) source is used.
1
Color Pattern. This bit is decoded from the last instruction’s opcode field to identify whether a color (1)
pattern is used.
0
Monochromatic Pattern. This bit is decoded from the last instruction’s opcode field to identify whether a
monochrome (1) pattern is used.
221
®
12.4.2
R
BR01—Setup BLT Raster OP, Control, and Destination Offset
Default:
Attributes:
40004h
0000 xxxx
BR01 contains the contents of the last Setup instruction dword 1. It is identical to the BLT Raster OP, Control, and
Destination Offset definition, but it is used with the following instructions: PIXEL_BLT, SCANLINE_BLT, and
TEXT_BLT.
31
30
29
28
27
Sol
Pat
Rsvd
Mono
Src
Trans
Mono
Pat
Trans
Src
Sel
Mode
15
14
Rsvd
26
25
Dyn
Color
Depth
Enable
24
23
Color
Depth
13
16
Raster Operation
0
Destination Pitch (Offset)
Bit
Descriptions
31
Solid Pattern Select. This bit applies only when the pattern data is monochromatic. This bit determines
whether or not the BLT engine actually performs read operations from the frame buffer in order to load the
pattern data. Use of this feature to prevent these read operations can increase the BLT engine
performance, if use of the pattern data is indeed unnecessary. The BLT engine is configured to accept
either monochromatic or color pattern data via the opcode field.
0 = This causes normal operation with regard to the use of the pattern data. The BLT engine proceeds with
the process of reading the pattern data, and the pattern data is used as the pattern operand for all bitwise operations.
1 = The BLT engine forgoes the process of reading the pattern data. (It is presumed that all bits of pattern
data are set to 0.) The pattern operand for all bit-wise operations is forced to the background color
specified in the Color Expansion Background Color Register.
30
Reserved
29
Monochrome Source Transparency Mode. This bit applies only when the source data is monochromatic.
This bit determines whether or not the byte(s) at the destination corresponding to the pixel to which a given
bit of the source data also corresponds will actually be written, if that source data bit has the value of 0.
This feature can make it possible to use the source as a transparency mask. The BLT engine is configured
to accept either monochromatic or color source data via the opcode field.
0 = This causes normal operation with regard to the use of the source data. Wherever a bit in the source
data has the value of 0, the color specified in the background color register is used as the source
operand in the bit-wise operation for the pixel corresponding to the source data bit, and the bytes at the
destination corresponding to that pixel are written with the result.
1 = Wherever a bit in the source data has the value of 0, the byte(s) at the destination corresponding to the
pixel to which the source data bit also corresponds are simply not written, and the data at those byte(s)
at the destination are left unchanged.
222
®
R
Bit
Descriptions
28
Monochromatic Pattern Transparency Mode. This bit applies only when the pattern data is
monochromatic. This bit determines whether or not the byte(s) at the destination corresponding to the pixel
to which a given bit of the pattern data also corresponds will actually be written, if that pattern data bit has
the value of 1. This feature can enable the use of the pattern as a transparency mask. The BLT engine is
configured to accept either monochromatic or color pattern data via the opcode field.
0 = This causes normal operation with regard to the use of the pattern data. Wherever a bit in the pattern
data has the value of 0, the color specified in the background color register is used as the pattern
operand in the bit-wise operation for the pixel corresponding to the pattern data bit, and the bytes at the
1 = Wherever a bit in the pattern data has the value of 0, the byte(s) at the destination corresponding to the
pixel to which the pattern data bit also corresponds are simply not written, and the data at those byte(s)
27
Source Select Mode. Configures the BLT engine data source.
0 = The BLT engine reads the source data from the frame buffer at the location specified in the Source
Address Register.
1 = The BLT engine accepts the source data from instruction stream controller through the
IMMEDIATE_BLT instruction. The BLT engine will hang if it doesn't get an even number of DWs.
26
Dynamic Color Depth Enable.
0 = Use BLTCNTL Register (offset 7000Ch) for color depth specification.
1 = Use bits [25:24] for color depth specification. This supersedes the color depth specified by the
BLTCNTL Register.
25:24
Dynamic Color Depth.
00 = 8-bit color depth
23:16
Raster Operation Select. These 8 bits are used to select which one of 256 possible raster operations is to
be performed by the BLT engine. The 8-bit values and their corresponding raster operations are intended to
correspond to the 256 possible raster operations specified for graphics device drivers in the Microsoft*
Windows* environment. The opcode field must indicate a monochrome source if ROP = F0.
15:14
Reserved
13:0
Destination Pitch (Offset). These 14 bits store the signed memory address offset value that is used to
increment or decrement the destination address originally specified in the Destination Address Register, as
each scan line’s worth of destination data is written into the frame buffer by the BLT engine, so that the
destination address will point to the next memory address to which the next scan line’s amount of
destination data is to be written.
If the intended destination of a BLT operation is within on-screen frame buffer memory, this offset is
normally set so that each subsequent scan line’s worth of destination data lines up vertically with the
destination data in the previous scan line. However, if the intended destination of a BLT operation is within
off-screen memory, this offset can be set so that each subsequent scan line’s worth of destination data is
stored at a location immediately after the location where the destination data for the last scan line ended, in
order to create a single contiguous block of bytes of destination data at the destination.
223
®
12.4.3
R
BR02—Clip Rectangle Y1 Address
Default:
Attributes:
40008h
None
BR02 is loaded by either the SETUP_BLT or SETUP_MONO_PATTERN_SL_BLT instruction and is used with the
PIXEL_BLT, SCANLINE_BLT or TEXT_BLT instruction.
31
26
25
Reserved
0
Clip Rectangle Y1 Address Bits [25:0]
Bit
Descriptions
31:26
Reserved. The maximum GC graphics address is 64 MB. Debug implementation specific = Leftdiscard[5:0]
25:0
Clip Rectangle Y1 Address Bits [25:0]. These 26 bits specify the top clipping address of the destination
data. This clip compare is inclusive (i.e., draw if destination address is greater than or equal to). This
address points to the first address of a scan line. The Clip Rectangle X registers take care of the pixel
positions within a scan line.
12.4.4
BR03—Clip Rectangle Y2 Address
Default:
Attributes:
4000Ch
None
BR03 is loaded by either the SETUP_BLT or SETUP_MONO_PATTERN_SL_BLT instructions and is used with
31
26
Reserved
Bit
25
0
Clip Rectangle Y2 Address Bits [25:0]
Descriptions
31:26
Reserved. The maximum GC graphics address is 64 MB. Debug implementation specific =
Leftdiscard[11:06]
25:0
Clip Rectangle Y2 Address Bits [25:0]. These 26 bits specify the bottom clipping address of the
destination data. This clip compare is inclusive (i.e., draw if destination address is less than or equal to).
This address points to the first byte of a scan line. The Clip Rectangle X registers take care of the pixel
positions within a scan line.
224
®
R
12.4.5
BR04—Clip Rectangle X1 and X2
Default:
Attributes:
40010h
None
BR04 is loaded by either the SETUP_BLT or SETUP_MONO_PATTERN_SL_BLT instructions and is used with
31
28
27
Reserved
15
16
Clip Rectangle X2 coordinate (right) [11:00]
12
11
Reserved
0
Clip Rectangle X1 coordinate (left) [11:00]
Bit
Descriptions
31:28
Reserved.
27:16
Clip Rectangle X2 Coordinate. These 12 bits specify the rightmost X coordinate that is written to the
destination. The comparison is inclusive, with a less than or equal. The byte address of this coordinate is:
scan line address + X2 * bytes/pixel
15:12
Reserved.
11:0
Clip Rectangle X1 Coordinate. These 12 bits specify the leftmost X coordinate which is written to the
destination. The comparison is inclusive, with a greater than or equal. The byte address of this coordinate
is:
scan line address + X1 * bytes/pixel
12.4.6
BR05—Setup Expansion Background Color
Default:
Attributes:
31
40014h
None
24
Reserved
Bit
23
0
Setup Expansion Background Color Bits [23:0]
Descriptions
31:24
Reserved.
23:0
Setup Expansion Background Color Bits [23:0]. These bits provide the one, two or three bytes of color
data that select the background color to be used in the color expansion of monochrome pattern or source
data for either the SCANLINE_BLT or TEXT_BLT instructions. BR05 is also used as the solid pattern for
the PIXEL_BLT instruction.
Whether one, two or three bytes of color data are needed depends upon the color depth to which the BLT
engine has been set. For a color depths of 24 bpp, 16 bpp and 8 bpp, bits [23:0], [15:0] and [7:0],
respectively, are used.
225
®
12.4.7
R
BR06—Setup Expansion Foreground Color
Default:
Attributes:
31
40018h
None
24
23
0
Reserved
Setup Expansion Foreground Color Bits [23:0]
Bit
Descriptions
31:24
Reserved.
23:0
Setup Expansion Foreground Color Bits [23:0]. These bits provide the one, two or three bytes of color
data that select the foreground color to be used in the color expansion of monochrome pattern or source
data for either the SCANLINE_BLT or TEXT_BLT instructions.
engine has been set. For a color depth of 24 bpp, 16 bpp and 8 bpp, bits [23:0], [15:0] and [7:0],
12.4.8
BR07—Setup Color Pattern Address
Default:
Attributes:
31
4001Ch
None
26
25
16
Reserved
Setup Color Pattern Address Bits [25:16]
15
6
Setup Color Pattern Address Bits [15:6]
Bit
5
0
Reserved
Descriptions
31:26
Reserved. The maximum GC graphics address is 64 MB.
25:6
Pattern Address. These 20 bits specify the starting address of the color pattern from the SETUP_BLT
instruction. This register works identically to the Pattern Address register, but this version is only used with
the SCANLINE_BLT instruction execution.
The pattern data must be located on a pattern-size boundary. The pattern is always of 8x8 pixels.
Therefore, its size depends upon its pixel depth. The pixel depth may be 8, 16 or 24 bits per pixel, if the
pattern is in color. (The pixel depth of a color pattern must match the pixel depth to which the graphics
system has been set.) Monochromatic patterns require 8 bytes and are supplied through the instruction.
Color patterns of 8-, 16-, and 24-bits-per-pixel color depth must start on 64-, 128-, and 256-byte
boundaries, respectively.
Note:
In the case of 24 bits per pixel, each scan line (i.e., each row of 8 pixels) of pattern data takes up 24
®
consecutive bytes, not 32 like the Intel 740 Graphics Controller.
5:0
226
Reserved. These bits always return 0 when read.
®
R
12.4.9
BR08—Destination X1 and X2
Default:
Attributes:
40020h
None
BR08 is loaded by either the PIXEL_BLT, SCANLINE_BLT or TEXT_BLT instruction. The PIXEL_BLT instruction
only writes the destination X coordinate register.
31
28
27
Reserved
15
12
Reserved
16
Destination X2 coordinate (right) [11:00]
11
0
Destination X1 or X coordinate (left) [11:00]
Bit
Descriptions
31:28
Reserved.
27:16
Destination X2 coordinate. These 12 bits specify the rightmost X coordinate that is written to the
destination if not clipped. The comparison is inclusive, with a less than or equal to. The byte address of this
coordinate is:
scan line address + X2 * bytes/pixel.
15:12
Reserved
11:0
Destination X1 or X coordinate (left). These 12 bits specify the leftmost X coordinate that is written to the
destination. This is also the working register, where it changes while the BLT engine is working. The
comparison is inclusive, with a greater than or equal to. The byte address of this coordinate is:
scan line address + X1 * bytes/pixel.
Note:
Some instructions affect only one pixel, requiring only one coordinate. Other instructions affect multiple
pixels and need both coordinates.
227
®
R
12.4.10 BR09—Destination Address and Destination Y1 Address
Default:
Attributes:
31
26
40024h
None
25
Reserved
0
Destination and Destination Y1 and Y Address Bits [25:0]
Bit
Descriptions
31:26
Reserved.
25:0
Destination and Destination Y1 and Y Address Bits. These 26 bits specify the starting pixel address of
the destination data. This register is also the working destination address register and changes as the BLT
engine performs accesses.
The following are used as the scan line address (destination Y address & destination Y1 address) for BLT
instructions: PIXEL_BLT, SCANLINE_BLT, and TEXT_BLT. In this case the address points to the first pixel
in a scan line, and it is compared with the ClipRect Y1 & Y2 address registers to determine whether or not
the scan line should be written. The Destination Y1 address is the top scan line to be written for text.
This register is always the last register written for a BLT drawing instruction. Writing BR09 starts the BLT
engine execution.
Note:
Some instructions affect only one scan line, requiring only one coordinate. Other instructions affect multiple
scan lines and need both coordinates.
Note:
This is a working register. If BR09 is read while the BLT engine is busy, the contents will be unpredictable.
12.4.11 BR10—Destination Y2 Address
Default:
Attributes:
31
26
Reserved
Bit
40028h
None
25
0
Destination Y2 Address Bits [25:0]
Descriptions
31:26
Reserved. The maximum GC graphics address is 64 MB. Debug implementation specific = bmtdpix[5:0]
25:0
Destination Y2 Address. The following is used as the scan line address (destination Y2 address) for BLT
instruction: TEXT_BLT. The address points to the first pixel in a scan line, and it is compared with the
ClipRect Y1 & Y2 address registers to determine whether or not the scan line should be written. The
destination Y2 address is the bottom scan line to be written for text.
228
®
R
12.4.12 BR11—BLT Source Pitch (Offset) or Monochrome Source Quadwords
Default:
Attributes:
4002Ch
None
31
16
BLT Engine Status – TBS
Bit
15
14
Reserved
13
0
Source Pitch (Offset) or Monochrome
Source Quadwords
Descriptions
31:16
BLT Engine Status. This field is used to read back an important debug status. It will be specified in the
future.
15:14
Reserved
13:0
Source Pitch (Offset) or Monochrome Source Quadwords. When the color source data is located within
the frame buffer or AGP aperture, these signed 14 bits store the memory address offset (pitch) value by
which the source address originally specified in the Source Address Register is incremented or
decremented, as each scan line’s worth of source data is read from the frame buffer by the BLT engine, so
that the source address will point to the next memory address from which the next scan line’s worth of
source data is to be read.
When the source data is provided by IMMEDIATE_BLT instruction, these 14 bits store the number of bytes
to be counted, from the beginning of a scan line’s worth of data to where the next scan line’s worth of data
begins.
Note that if the intended source of a BLT operation is within on-screen frame buffer memory, this offset is
normally set to accommodate the fact that each subsequent scan line’s worth of source data lines up
vertically with the source data in the previous scan line. However, if the intended source of a BLT operation
is within off-screen memory, this offset can be set to accommodate a situation in which the source data
exists as a single contiguous block of bytes wherein each subsequent scan line’s worth of source data is
stored at a location immediately after the location where the source data for the last scan line ended.
When monochrome source data is being processed, this field is used for indicating the total number of
quadwords of data in the source data stream.
229
®
R
12.4.13 BR12—Source Address
Default:
Attributes:
31
26
40030h
None
25
0
Reserved
Source Address Bits [25:0]
Bit
Descriptions
31:26
Reserved. The maximum GC Graphics address is 64 MB.
25:0
Source Address Bits [25:0]. These 26 bits are used to specify the starting pixel address of the color
source data. The lower 3 bits are used to indicate the position of the first valid byte within the first quadword
of the source data.
Note:
This is a working register. If BR12 is read while the BLT engine is busy, the contents will be unpredictable. The value
contained in this register while the BLT engine is busy is unrelated to the programmed value that can be read when
the BLT engine is idle.
12.4.14 BR13—BLT Raster OP, Control, and Destination Pitch
Default:
Attributes:
40034h
0000 xxxx
31
30
29
28
27
26
Sol
Pat
Sel
X
Dir
Mon
Src
Trans
Mon
Pat
Trans
Src
Sel
Mode
Dyn
Col
En
15
14
Reserved
25
24
23
Dynamic
Color
Depth
16
Raster Operation Select
13
0
Destination Pitch (Offset)
Bit
Descriptions
31
Solid Pattern Select. This bit applies only when the pattern data is monochromatic. This bit determines
whether or not the BLT engine actually performs read operations from the frame buffer in order to load the
pattern data. The use of this feature to prevent these read operations can increase the BLT engine
performance, if the use of the pattern data is indeed unnecessary. The BLT engine is configured to accept
either monochromatic or color pattern data via the opcode field.
0 = This causes normal operation with regard to the use of the pattern data. The BLT engine proceeds with
the process of reading the pattern data, and the pattern data is used as the pattern operand for all bitwise operations.
1 = The BLT engine forgoes the process of reading the pattern data. (It is presumed that all of the bits of
the pattern data are set to 0.) The pattern operand for all bit-wise operations is forced to the
background color specified in the Color Expansion Background Color Register.
230
®
R
Bit
30
Descriptions
X Increment/Decrement Select.
0 = The bytes corresponding to the pixels within each scan line of data are written to the destination, with
the bytes corresponding to the left-most pixel of each scan line at the destination being the first to be
written, and then proceeding rightward toward the rightmost pixel.
1 = The bytes corresponding to the pixels within each scan line of data are written to the destination, with
the bytes corresponding to the rightmost pixel of each scan line at the destination being the first to be
written, and then proceeding in a leftward direction toward the leftmost pixel.
29
Monochrome Source Transparency Mode. This bit applies only when the source data is in
to which a given bit of the source data also corresponds will actually be written if that source data bit has
the value of 0. This feature can enable the use of the source as a transparency mask. The BLT engine is
configured to accept either monochromatic or color source data via the opcode field.
0 = This causes normal operation with regard to the use of the source data. Wherever a bit in the source
data has the value of 0, the color specified in the background color register is used as the source
operand in the bit-wise operation for the pixel corresponding to the source data bit, and the bytes at the
1 = Where a bit in the source data has the value of 0, the byte(s) at the destination corresponding to the
pixel to which the source data bit also corresponds simply are not written, and the data at those byte(s)
28
27
26
Monochromatic Pattern Transparency Mode. This bit applies only when the pattern data is
to which a given bit of the pattern data also corresponds will actually be written, if that pattern data bit has
the value of 1. This feature can make it possible to use the pattern as a transparency mask. The BLT
engine is configured to accept either monochromatic or color pattern data via the opcode in the Opcode
and Control register.
0
This causes normal operation with regard to the use of the pattern data. Where a bit in the pattern data
has the value of 0, the color specified in the background color register is used as the pattern operand in
the bit-wise operation, for the pixel corresponding to the pattern data bit, and the bytes at the
1
Wherever a bit in the pattern data has the value of 0, the byte(s) at the destination corresponding to the
pixel to which the pattern data bit also corresponds simply are not written, and the data at those byte(s)
Source Select Mode.
0
Configures the BLT engine to read the source data from the frame buffer at the location specified in the
Source Address Register.
1
Configures the BLT engine to accept the source data from the instruction stream controller through the
IMMEDIATE_BLT instruction. The BLT engine will hang if it doesn't get an even number of DWs.
Dynamic Color Depth Enable.
0 = Use BLTCNTL Register (offset 7000Ch) for color depth specification.
1 = Use bits [25:24] for color depth specification. This supersedes the color depth specified by BLTCNTL
Register.
25:24
Dynamic Color Depth.
11 = Reserved
231
®
R
Bit
Descriptions
23:16
Raster Operation Select. These 8 bits are used to select which one of 256 possible raster operations is to
be performed by the BLT engine. The 8-bit values, and their corresponding raster operations, are intended
to correspond to the 256 possible raster operations specified for graphics device drivers in the Microsoft*
Windows* environment. The opcode must indicate a monochrome source operand if ROP = F0.
15:14
Reserved
13:0
Destination Pitch (Offset). These 14 bits store the signed memory address offset value by which the
destination address originally specified in the Destination Address Register is incremented or decremented,
as each scan line’s worth of destination data is written into the frame buffer by the BLT engine, so that the
destination address will point to the next memory address to which the next scan line’s worth of destination
data is to be written.
If the intended destination of a BLT operation is within on-screen frame buffer memory, this offset is
normally set so that each subsequent scan line’s worth of destination data lines up vertically with the
destination data in the scan line above. However, if the intended destination of a BLT operation is within offscreen memory, this offset can be set so that each subsequent scan line’s worth of destination data is
stored at a location immediately after the location where the destination data for the last scan line ended, in
order to create a single contiguous block of bytes of destination data at the destination.
12.4.15 BR14—Destination Width & Height
Default:
Attributes:
40038h
None
BR14 contains the values for the height and width of the data to be subjected to BLT. If these values are incorrect, such
that the BLT engine either expects data it does not receive or receives data it did not expect, the system can hang.
31
29
28
Reserved
15
Bit
31:29
28:16
Destination Height
13
Reserved
16
12
0
Destination Byte Width
Descriptions
Reserved
Destination Height. These 13 bits specify the height of the destination data, in terms of the number of
scan lines. This is a working register.
15:13
Reserved
12:0
Destination Byte Width. These 13 bits specify the width of the destination data, in terms of the number of
bytes per scan line. The number of pixels per scan line into which this value translates depends upon the
color depth to which the graphics system has been set.
Note:
This is a working register. If BR14 is read while the BLT engine is busy, the contents will be unpredictable. The value contained
in this register while the BLT engine is busy is unrelated to the programmed value that can be read while the BLT engine is idle.
232
®
R
12.4.16 BR15—Color Pattern Address
Default:
Attributes:
31
4003Ch
None
26
25
16
Reserved
Color Pattern Address Bits [25:16]
15
6
Color Pattern Address Bits [15:6]
Bit
5
0
Reserved
Descriptions
31:26
25:6
Color Pattern Address. These 20 bits specify the starting address of the pattern.
The pattern data must be located on a pattern-size boundary. The pattern is always of 8x8 pixels.
Therefore, its size depends upon its pixel depth. The pixel depth may be 8, 16 or 24 bits per pixel, if the
pattern is in color. (The pixel depth of a color pattern must match the pixel depth to which the graphics
system has been set.) Monochromatic patterns require 8 bytes and are supplied through the instruction.
Color patterns of 8-, 16-, and 24-bits-per-pixel color depth must start on 64-byte, 128-byte, and 256-byte
boundaries, respectively.
Note:
In the case of 24 bits per pixel, each scan line’s worth (i.e., each row of 8 pixels) of pattern data takes
up 32 consecutive bytes, not 24. It is formatted so that there is a contiguous block of 8 sets of 3 bytes,
each corresponding to 1 of the 8 pixels, followed by a contiguous block of the 8 extra bytes. When the
BLT engine reads 24-bits-per-pixel pattern data, it will read only the first 24 bytes of each scan line’s
worth of data, picking up the 8 sets of 3 bytes for each of the 8 pixels, and entirely ignoring the
remaining 8 bytes.
5:0
Reserved. These bits always return 0 when read.
233
®
R
12.4.17 BR16—Pattern Expansion Background & Solid Pattern Color
Default:
Attributes:
31
40040h
None
24
23
Reserved
0
Pattern Expansion Background Color Bits [23:0]
Bit
Descriptions
31:24
Reserved.
23:0
Pattern Expansion Background Color Bits [23:0]. These bits provide the one, two, three or four bytes of
color data that select the background color to be used in the color expansion of monochrome pattern data
during BLT operations.
Whether one, two, three or four bytes of color data are needed depends upon the color depth to which the
BLT engine has been set. For a color depth of 24 bpp, 16 bpp, and 8 bpp, bits [23:0], [15:0], and [7:0],
12.4.18 BR17—Pattern Expansion Foreground Color
Default:
Attributes:
31
40044h
None
24
Reserved
Bit
23
0
Pattern Expansion Foreground Color Bits [23:0]
Descriptions
31:24
Reserved
23:0
Pattern Expansion Foreground Color Bits [23:0]. These bits provide the one, two, three or four bytes of
color data that select the foreground color to be used in the color expansion of monochrome pattern data
Whether one, two or three or bytes of color data are needed depends upon the color depth to which the
234
®
R
12.4.19 BR18—Source Expansion Background, and Destination Color
Default:
Attributes:
31
40048h
None
24
23
Reserved
0
Source Expansion Background Color Bits [23:0]
Bit
Descriptions
31:24
Reserved. Debug implementation specific [31:25] = bmnewcliplw[6:0].
23:0
Source Expansion Background Color Bits [23:0]. These bits provide the one, two or three bytes of color
data that select the background color to be used in the color expansion of monochrome source data during
BLT operations.
This register is also used to support destination transparency mode and solid color fill.
Whether one, two, three or four bytes of color data are needed depends upon the color depth to which the
12.4.20 BR19—Source Expansion Foreground Color
Default:
Attributes:
31
24
Reserved
Bit
4004Ch
None
23
0
Source Expansion Foreground Color Bits [23:0]
Descriptions
31:24
Reserved. Debug implementation specific [31:26] = bmnewcliplw[12:07].
23:0
Pattern/Source Expansion Foreground Color Bits [23:0]. These bits provide the one, two or three bytes
of color data that select the foreground color to be used in the color expansion of monochrome source data
engine has been set. For a color depth of 24 bpp, 16 bpp, and 8 bpp, bits [23:0], [15:0], and [7:0],
235
®
R
12.4.21 S_SLADD—Source Scan Line Address
Default:
Attributes:
31
26
40074h
None
25
0
Reserved
Source Scan Line Address Bits [25:0]
Bit
Descriptions
31:26
25:0
Source Scan Line Address. These 26 bits are used by the overscan line fetching state machine to
address the source data. Source data is read when ever there is room in the internal memory buffer to
avoid the read latency between scan lines.
Note:
This is a working register. If this register is read while the BLT engine is busy, the contents will be unpredictable.
12.4.22 D_SLH—Destination Scan Line Height
Default:
Attributes:
40078h
None
D_SLH is a working register that contains the current height used by the overscan line state machine.
31
16
15
Destination Scan Line Height
Bit
0
Reserved
Descriptions
31:16
Destination Scan Line Height. These 16 bits specify the height of the destination data, in terms of the
number of scan lines.
15:0
Reserved
Note:
236
®
R
12.4.23 D_SLRADD—Destination Scan Line Read Address
Default:
Attributes:
31
26
Reserved
Bit
4007Ch
None
25
0
Destination Scan Line Read Address Bits [25:0]
Descriptions
31:26
25:0
Destination Address. These 26 bits specify the scan line address of the destination data being read. This
is a working register.
Note:
237
®
R
238
®
R
13. Rendering Engine Instructions
This chapter describes the 3D instructions that manage the Graphics Controller (GC) rendering engine. The GC
rendering engine receives all software driver instructions through the instruction interface (ring buffers).
13.1 GFXPRIMITIVE
This instruction performs most of the rendering operations handled by the GC. Triangle and triangle lists are rendered
using this instruction. Triangle strips and fans reduce the number of vertices that must be delivered to the hardware for
adjacent triangles and that also are supported by this instruction. Lines and line lists continue to be supported. The GC
also supports the rendering of axis-aligned rectangles. The vertices of these primitives may be of variable length. For
example, one call to GFXPRIMITIVE may include only the X, Y values at the vertices, while another call may specify
all forty-four bytes of attribute information at each vertex. The following sections consider the primitive in detail.
V2
V1
V0
Figure 33. Rectangle vertices
The GC supports both D3D and OpenGL notations. The selection can be done in the
GFXRENDERSTATE_RASTER_RULE state variable.
13.1.1
Axis-Aligned Rectangles
Axis-aligned rectangles are described by three vertices, as shown in the adjacent figure. The vertices should always
describe a right triangle, such that the base of the triangle is parallel to the x-axis and the vertical leg of the triangle is
parallel to the y-axis.
13.1.2
Primitive Winding Order
The winding order of the vertices is considered if primitive culling is enabled. In the preceding example, the winding
order is clockwise. In a triangle strip, the winding order toggles on every triangle (e.g., CW, CCW, CW, CCW, etc.).
Therefore, hardware toggles the culling orientation on every triangle to match up with the strip sequence. If the primitive
type is triangle strip, culling orientation is toggled on the 2nd, 4th, and 6th triangles, and so on. If the primitive type is
triangle strip with reverse winding order, culling orientation is toggled on the 1st, 3rd, and 5th triangles, and so on.
239
®
13.1.3
R
Position Mask
In the variable vertex format, a position mask is sent to indicate the presence of the X, Y, Z, and 1/W parameters. If 1/W
(RHW) is declared not present in the vertex packet, hardware forces 1/w equal to 1. In the case Z is declared not present,
the Z value of the last triangle from the last vertex instruction packet will be used for the triangles in this packet.
13.1.4
Bias
For each polygon, a Z bias value can be included in the vertex instruction packet. This value is a floating-point number
that ranges from -1 to 1. The Z bias value is added to the Z value and then clamped to between 0 to 1, before primitive
setup calculation. In the triangle or line primitive type, the Z bias of the last vertex (3rd in triangle, 2nd in line) is used to
add to all the vertices of the primitive, and the Z bias value on the other vertices is ignored. In triangle strip and fan, the
Z bias value on the 3rd vertex will be used for the first triangle, and the Z bias value on a subsequent vertex that defines a
triangle will be used for that triangle.
13.2 Primitive Rendering Instruction Format
DWord
Bits
0
31:29
Client: 03h – Rendering Engine
28:24
Rendering Primitive: 1Fh
23
1
Reserved: 00h
22:18
Primitive Type:
0h – Triangle or triangle List
1h – Triangle strip
2h – Triangle strip with reverse winding order
3h – Triangle fan
4h – Polygon (triangle fan with common flat-shaded vertex)
5h – Line or line list
6h – Line strip
7h – Rectangle or rectangle list (must be axis aligned)
8h-1Fh – Reserved
17:0
Dword Count
N/A
Primitive[0] Vertex[0]: GFXVERTEX
N/A
Primitive[0] Vertex[1]: GFXVERTEX
…
Variable
Description
N/A
…
Primitive[n] Vertex[2]: GFXVERTEX
Last DWORD of last vertex of last primitive
Refer to the GFXVERTEX section for details on defining a primitive vertex.
240
®
R
13.2.1
Variable-Length Vertex Formats for Rendering Instructions
The GC supports variable-length vertex formats. These formats are determined by enable bits contained in the variablelength vertex format (VLVF) instructions. The following table specifies the attributes associated with each vertex. The
order of the attributes must be strictly observed.
For any attribute that is defined as not present, the value of the corresponding attribute from the last triangle of the last
packet will be used for the triangles in this packet. However, in the case of strip and fan primitives, since the vertex
numbers rotate from one primitive to the next primitive (e.g., the third vertex of one triangle becomes the first vertex of
the next triangle), the reused attribute value from the last primitive is not guaranteed to be on the same vertex of the next
primitive. In this case, undesired results may occur. It is recommended to turn off features that have the corresponding
attributes disabled. (For example, if no texture coordinate is sent, then turn off the texture mapping feature.)
Vertex Attribute
Comments
X Position
Required for every vertex
Triangle Edge V2-V0 Enable
Required only on the first vertex of a triangle and on each new vertex, which
defines a triangle when processing strips and fans.
Y Position
Required for every vertex
Z Position
Optional
Z Bias
Optional
Reciprocal of W
Optional
Diffuse Color (Alpha, Red, Green
and Blue)
Optional
Specular Color/Fog Factor
Optional
Texture Coordinate Set 0
Optional
Texture Coordinate Set 1
Optional
241
®
R
13.3 GFXVERTEX
Rendering engine instructions include the data per vertex. The vertex data format is the same for these instructions.
GFXVERTEX is not an instruction, but a definition of this vertex data format. GFXVERTEX does not include an
instruction header since it is not an instruction. The GFXVERTEX format is as follows:
DWord
Bits
Description
0
31:4
X position: Relative to origin drawing rectangle or dest buffer. Valid data range is -383 to 1663.
(IEEE float, except 19-bit mantissa)
Reserved: 00h
2
Triangle Edge V2-V0 Enable: This flag enables antialiasing and wireframe capabilities for the
triangle edge defined by the third and first vertices (1 = enable, 0 = disable). This flag field is
valid only in the first vertex of the triangle definition. It is reserved otherwise.
1
triangle edge defined by the second and third vertices (1 = enable, 0 = disable). This flag field is
0
triangle edge defined by the first and second vertices (1 = enable, 0 = disable). This flag field is
1
31:0
Y position: Relative to origin of drawing rectangle or dest buffer. Valid data range is -383 to
1663. (IEEE float)
2
31:0
Z position: Normalized depth. Valid data range is 0 to 1. (IEEE float)
3
31:0
Z Bias: Valid data range is -1 to 1. (IEEE float)
4
31:0
Reciprocal of W: Valid data is any positive number.(IEEE float)
5
31:24
Color Alpha: Valid data range is 0 to 255. (unsigned int)
23:16
Color Red: Valid data range is 0 to 255. (unsigned int)
15:8
Color Green: Valid data range is 0 to 255. (unsigned int)
7:0
Color Blue: Valid data range is 0 to 255. (unsigned int)
31:24
Fog Factor: Valid data range is 0 to 255. (unsigned int)
23:16
Specular Red: Undefined
15:8
Specular Green: Undefined
7:0
Specular Blue: Valid data range is 0 to 255. (unsigned int)
7
31:0
Tu 0: Texture coordinates. Data valid over entire IEEE floating-point range. (IEEE float)
8
31:0
TV 0: Texture coordinates. Data valid over entire IEEE floating-point range. (IEEE float)
9
31:0
Tu 1: Texture coordinates. Data valid over entire IEEE floating-point range. (IEEE float)
10
31:0
TV 1: Texture coordinates. Data valid over entire IEEE floating-point range. (IEEE float)
6
242
3
®
R
13.4 GFXRENDERSTATE_VERTEX_FORMAT
Flexible Vertex Format Packet
DWord
Bit
Description
0
31:29
28:24
3DState24: 05h
23:12
Reserved: 00h
11:8
Texture Coordinate Count: This field identifies how many coordinates are present in the vertex.
The valid range is 0-2.
7
Specular Color and Fog Factor Present
6
Diffuse Color and Alpha Present
5
Z-Offset Present
4
Reserved: 00h (DX6 Normal Vector Present Bit)
3:1
0
Position Mask:
0h – Invalid
1h – XYZ present, RHW not present
2h – XYZRHW present
3h – XY present, RHW and Z not present
4h – XYRHW present, Z not present
5h-7h – Reserved
Reserved: 00h
243
®
13.4.1
R
Non-Pipelined State Variables
The Intel® 82810 Chipset does not optimize on state variables that change infrequently. For performance reasons, ISVs
are asked to group polygons for state changes. In the Intel® 82810 Chipset, the following state variables will not be
pipelined:
Z_BIAS[7:0], MONO[1:0], FOG_CLR[23:0], ALPHA_REF[7:0], ALPHA_FUNC[2:0], COLOR_KEYH[15:0],
COLOR_KEYL[15:0], COLOR_INDEX[7:0], COLORKEY_EN, CHROMKEY_EN,
DEST_BUFFER_INFO, DRAWING_RECTANGLE_INFO, SCISSOR_RECTANGLE_INFO
These state variables have been grouped appropriately into packets that tell the instruction parser that these are nonpipelined state variables. The 3D pipeline will be flushed up to and including the Color Calculator stage before these
state variables are updated. The pixel cache or the streamers do not need to be flushed for state change. It is important for
the application to group these changes into one pipeline to minimize performance impact. This optimization has no
software impact other than a slight performance impact.
13.5 GFXRENDERSTATE_MAP_TEXELS
The mapping engine is capable of generating at most two texels per pixel. The texels may be obtained from two separate
maps or the same map using different u,v coordinates. The binding between the texels that are generated by the engine
and the coordinate set and the map information state is specified with this instruction. The texels are generated in the
order specified by the Texel Index. The texture mapping is disabled by default or if no texel is enabled. If one of the two
texels is enabled, one texture coordinate will be used for texture mapping. If both texels are enabled, both texture
coordinates will be used for texture mapping.
Coordinate Sets
Map Information
Texels
Blending Stages
Figure 34. State variable relationships
244
®
R
DWord
Bit
Description
0
31:29
28:24
3DState16: 1Ch
23:19
Opcode: 0h
18:16
Reserved: 00h (Additional Texels)
15
Texel 1 State Variable Mask (0: do not update; 1: update)
14
Texel 1 Enable:
0 – Disable (default)
1 – Enable
13:12
11
10:9
Reserved: 00h (Additional coordinate sets)
Texel 1 Coordinate Set Index: Index to the coordinate set
Reserved: 00h (Addition map information)
8
Texel 1 Map Information Index: Index to the map information
7
Texel 0 State Variable Mask (0: do not update; 1: update)
6
Texel 0 Enable:
0 – Disable (default)
1 – Enable
5:4
3
2:1
0
0
Reserved: 00h
Texel 0 Coordinate Set Index: Index to the coordinate set
Reserved: 00h
Texel 0 Map Information Index: Index to the map information
245
®
R
13.6 GFXRENDERSTATE_MAP_COORD_SETS
The mapping engine is capable of generating at most two map coordinate sets (u and v addresses) per pixel. Each output
texel may be related to separate coordinate sets or to the same coordinate set, as shown in the following figure. The
vertices of a GFXPRIMITIVE instruction may have 0, 1 or 2 map coordinate sets assigned, with varying settings
associated with each coordinate set.
A coordinate set may have normalized u/v coordinates or un-normalized coordinates. Normalized coordinates have been
divided by the size of the map prior to delivery to the GC. This means that map addresses, which lie on the map, have a
range of 0.0 to 1.0. 3D rendering utilizes normalized coordinates. Un-normalized coordinates are in units of texels/pixels
and have not been divided by the associated map’s height or width. Un-normalized coordinates allow the Arithmetic
Stretch Blitter operation to be specified in pixels avoiding floating-point issues. Un-normalized coordinates are not
usually specified outside of the map extents. These two coordinate sets may also have different wrap, clamp, and mirror
settings.
Coordinate Sets
Map Information
Texels
Blending Stages
246
®
R
DWord
Bit
Description
0
31:29
Client: 03h – Rendering engine
28:24
3DState16: 1Ch
23:19
Opcode: 1h
18:17
Reserved: 00h (Additional coordinate sets)
16
Update Coordinate Set Index: The valid range is 0-1.
15
Normalized Coordinate Set Mask (0: do not update; 1: update)
14
Normalized Coordinate Set:
0 – Coordinates are not normalized.
1 – Coordinates are normalized.
13:12
Reserved: 00h
11:8
Reserved: 00h (3-dimensional coordinates)
7
Address V State Variable Mask (0: do not update; 1: update)
6
Reserved: 00h
5:4
Address V: Valid values are:
0h – Wrap
1h – Mirror
2h – Clamp
3h – Wrap shortest
3
Address U State Variable Mask (0: do not update; 1: update)
2
Reserved: 00h
1:0
Address U: Valid values are:
0h – Wrap
1h – Mirror
2h – Clamp
3h – Wrap shortest
247
®
R
13.7 GFXRENDERSTATE_MAP_INFO
The mapping engine is capable of fetching texels from at most two maps per pixel. This instruction specifies the
attributes relating to the location and format of the map. Two different texels can be fetched from the same map.
Coordinate Sets
Map Information
Texels
Blending Stages
Discussed below are four classes of surface formats for the supported maps in the GC. The various attributes of the
texels/pixels are placed on four abstract channels named F0 -F3. These channels can be switched into four output
channels, which are delivered by the mapping engine to the color calculator. The switching of these channels (ME0 ME3) is controlled by the described Output Channel Selection field.
Table 10. Summary of Source Surface Formats with Filter Output Channel Mappings
#
0
Source Format
Bpt
Attribute Types and Formats
Filter Channels
F0
F1
F2
F3
I
I
I
Arbitrary Attribute
8
Alpha, intensity, luminance, chrominance, etc.
(I: 8)
I
1
Alpha Attribute
16
Alpha and luminance, chrominance, etc. (AY: 88)
Y
Y
T
A
2
Alpha Red Grn Blue
16
ARGB: 0565/1555/4444
R
G
B
A
3
4:2:2
YCrCb 4:2:2
Cr
Y
Cb
1’s
Certain memory tiling properties are supported directly by this instruction. This functionality is provided for surfaces
that do not need transparent CPU access (e.g., optimized texture maps). The following table identifies which fields are
used (✔) or ignored (⊗) for various permutations. The Pitch field must be specified in all cases.
Base
Address
Bits [31:26]
All Zeros
Utilize
Fence
Registers
Yes
Fence
Range
Hit
No
Tiled
Surface
Tile
Walk
Surface
⊗
⊗
Linear
All Zeros
Yes
Yes
⊗
Tiled
All Zeros
No
⊗
⊗
No
No
⊗
Yes
⊗
✔
Linear
All Zeros
*
Tiled
* The pitch specified in this instruction must be the same as the pitch in the corresponding fence register.
248
®
R
Surfaces that contain mip-maps are located within a single rectangular area of memory identified by the base address of
the upper-left corner and a pitch. The pitch must be specified at least as large as the next power of two that is equal to or
greater than the widest mip-map. These surfaces may be overlapped in memory and must adhere to the following
memory organization rules:
The base address must be 4-KB aligned.
Each successively smaller mip-map must lie vertically below and must be left aligned.
Each mip-map must have its upper-left corner vertically aligned with an even quadword address.
The following figures show an example of a 32×8 @ 16 bpt map and a 4×8 @ 16 bpt map, in which the dashed lines
identify quadwords.
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Base Address
32x8
16x4
8x2
4x1
2x1
1x1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
4x8
2x4
1x2
1x1
249
®
DWord
Bit
0
31:29
28:24
3DstateMW: 1Dh
23:16
Opcode: 0h
15:0
DWORD_LENGTH: 2h
31:29
Reserved: 0h
1
1
250
Description
28
Update Map Index: The valid range is 0-1.
27
Reserved: 0h
26:24
23
R
Surface Format:
0h – 8 bpt (indexed)
1h – 8 bpt
2h – 16 bpt
3h – Reserved (32 bpt)
4h – Reserved (4:2:0)
5h – 4:2:2
6h-7h – Reserved
Reserved: 0h
®
R
DWord
Bit
Description
1
22:21
Texel/Pixel Format:
8-bpt (Indexed) Surface Format
00 – RGB 565
01 – ARGB 1555
10 – ARGB 4444
11 – AY 88
8-bpt Surface Format
00 – Reserved
01 – Reserved
10 – Reserved
11 – Reserved
16-bpt Surface Format
00 – RGB 565
01 – ARGB 1555
10 – ARGB 4444
11 – AY 88
32-bpt Surface Format (not implemented in Intel® 82810 Chipset)
00 – ARGB 8888
01 – Reserved
10 – Reserved
11 – Reserved
4:2:0 (not implemented in Intel® 82810 Chipset)
00 – YcrCb Planar Format
01 – Reserved
10 – Reserved
11 – Reserved
4:2:2
00 – YcrCb, Swap Y format
01 – Reserved (YCrCb, Normal)
10 – Reserved (YCrCb, UV Swap)
11 – Reserved (YCrCb, UV/Y Swap)
1
20:19
Output Channel Selection: Selects the muxing for the possible 4 channels available in the surface
(at the output of the filter) and the 4 output channels from the mapping engine.
Mapping Engine Output Channels
ME0
ME1
ME2
ME3
0h –
F0
F1
F2
F3
1h –
⊗
F0
⊗
F3
2h –
⊗
F2
⊗
F3
3h – Reserved
18
Color Space Conversion Enable:
0 – Do not perform conversion.
1 – Perform color space conversion, assuming biased chrominance values.
251
®
DWord
R
Bit
17
Description
Vertical Line Stride: The number of lines to skip between logically adjacent lines.
0 – Do not skip any lines.
1 – Skip 1 line. (Provides support of interleaved/field surfaces.)
16
Vertical Line Stride Offset: The number of lines to add as an initial offset when the vertical line
stride is 1.
0 – Add no offset. (top field)
1 – Add 1. (bottom field)
15:11
10
Reserved: 0h
Utilize Fence Registers: If the registers are enabled, the Tiled Surface, Tile Height, Tile Walk
and Pitch are ignored. All request addresses are compared with the fence registers, and the
parameters associated with any matching fence register are utilized in the tiler. Otherwise, the
specified Tile Walk is utilized.
0 – Utilize Disable Fence Register
1 – Utilize Enable Fence Register
9
Tiled Surface: Specifies whether the surface is organized as rectangular memory or as tiled
memory. This field is ignored when the fence registers are being utilized.
0 – Linear (rectangular memory)
1 – Tiled
8
Tile Walk: The direction of increasing sequential addresses. This field is ignored when the fence
registers are being utilized or the surface is linear.
0 – X-Major
1 – Y-Major
7:5
Reserved: 0h
4
Reserved: 0h
1
3:0
Pitch: Log2 of the surface pitch is specified in quadwords. If the surface resides in a fenced
region, this value must correspond to the pitch specified when programming the fence
registers.
2
31
Dimensions are Powers of 2: This field specifies whether the following Height and Width fields of
the map are specified as the log2 of the actual dimension or as the actual height or width. If the
actual values are used, the coordinate set addresses must be non-normalized and are clamped.
(They cannot be wrapped or mirrored.)
0 – Height/width are the actual dimensions of the map, in pixels.
1 – Height/width are log2 of the actual dimensions of the base map.
30:26
Reserved: 00h
25:16
Height: If the dimensions are to be specified as powers of two (as determined by the Power of 2
field), this field must contain one of the following values:
0h – 1 texel high
4h – 16 texels high
ah – 1024 texels high
If the dimensions are to be specified as the actual values, this field contains the surface height –
1, in units of texels. The range of this value is 0-1023.
252
®
R
DWord
Bit
15:10
9:0
Description
Reserved: 00h
Width: If the dimensions are to be specified as powers of two (as determined by the Dimension
Power of 2 field), this field must contain one of the following values:
0h – 1 texel wide
4h – 16 texels wide
ah – 1024 texels wide
If the dimensions are to be specified as the actual values, this field contains the surface width –
1, in units of texels. The valid range for this value is 0-1023 for all surface formats, except 4:2:2.
For 4:2:2 surfaces, this value may be even 8-1022
3
31:4
3:0
Map Base Address: Virtual memory address, which is aligned on 16-byte boundaries. Bits [31:4]
of a 32-bit byte address are specified.
Reserved: 000h
253
®
R
13.8 GFXRENDERSTATE_MAP_FILTER
The mapping engine is capable of fetching texels/pixels from at most two maps per pixel. This instruction specifies the
filter settings associated with specified map.
Coordinate Sets
Map Information
Texels
Blending Stages
Anisotropic filtering produces superior image quality with reduced performance. Adjusting the LOD bias can contain the
aspect ratio of the filter. Generally the LOD bias should be set to -1.0 when anisotropic filtering is enabled. The Mip
Mode, Min Mode and Mag Mode state variables retain their meanings when anisotropic filtering is enabled.
The Mip Mode state variable specifies whether mip-mapping is enabled. If so, it also identifies whether the nearest mip
map should be used for the texture data or whether texels between the two nearest maps should be dithered together
when transitioning between maps.
The Min Mode state variable identifies the filtering operation to utilize when minimizing, and a texel is smaller than a
pixel. When this occurs, either nearest or linear filtering is performed. In nearest filtering, the texel with coordinates
nearest to the desired pixel value is used. In linear filtering, the weighted average of a 2-by-2 area of texels surrounding
the desired pixel is used.
The Mag Mode state variable describes operations where the map is magnified, and a texel is larger than one pixel. In
this mode, the finest texture map (LOD 0) is addressed, and the state variable selects between nearest and linear filtering.
In nearest filtering, the texel with coordinates nearest to the desired pixel value is used. In linear filtering, the weighted
average of a 2-by-2 area of texels surrounding the desired pixel is used.
254
®
R
DWord
Bit
0
31:29
28:24
3DState16: 1Ch
23:19
Opcode: 2h
18:17
Reserved: 0h
16
15:13
12
Description
Reserved: 0h
Anisotropic Filtering Enabled Mask:
0 – Do not update
1 – Update
11
Reserved: 0h
10
Anisotropic Filtering Enabled:
0 – Disable anisotropic filtering.
1 – Enable anisotropic filtering.
9
Mip Mode Filter Mask:
0 – Do not update
1 – Update
8
7:6
5
Reserved: 0h
Mip Mode Filter:
0 – None. Disable mip mapping.
1 – Nearest. Select the nearest mip map.
2, 3 – Reserved
Mag Mode Filter Mask:
0 – Do not update
1 – Update
4
Reserved: 0h
3
Mag Mode Filter:
0 – Nearest
1 – Linear
2
Min Mode Filter Mask:
0 – Do not update
1 – Update
0
1
Reserved: 0h
0
Min Mode Filter: Valid values are:
0 – Nearest
1 – Linear
255
®
R
13.9 GFXRENDERSTATE_MAP_LOD_LIMITS
The limits of the level-of-detail calculation can be controlled within the GC for each map. These values are specified in
the instruction, as follows.
Coordinate Sets
Map Information
Texels
Blending Stages
DWord
Bit
0
31:29
28:24
3Dstate16: 1Ch
23:19
Opcode: 3h
18:17
Reserved: 0h
16
15:14
13
Description
Reserved: 00h
Maximum Mip Level Mask:
0 – Do not update
1 – Update
12:5
4
Maximum Mip Level: This is the unsigned 8-bit value (4.4 format) that defines the highest
resolution map that may be accessed. A value of 0 equates to no limit on the mip map selection.
Values 1 through 11 will force the mip map selection to be between the specified value and the
highest defined mip map (i.e., the lowest-resolution map). This value must be less than the
minimum mip Level.
Minimum Mip Level Mask:
0 – Do not update
1 – Update
3:0
256
Minimum Mip Level: This is the unsigned 4-bit value that defines the lowest resolution map
that may be accessed. Values 0 through 10 will force the mip map selection to be between the
maximum resolution mip map and the specified value. This value must be correctly identified
for the current map. (No assumptions are made about the number of mips associated with a
particular map.) This value must be greater than the maximum mip level.
®
R
13.10 GFXRENDERSTATE_MAP_LOD_CONTROL
The level-of-detail bias can be associated with each map using the following instruction.
Coordinate Sets
Map Information
Texels
Blending Stages
DWord
Bit
0
31:29
28:24
3DState16: 1Ch
23:19
Opcode: 4h
18:17
Reserved: 00h
16
15:8
7
Description
Reserved: 00h
Texture LOD Bias Mask:
0 – Do not update
1 – Update
6:0
Texture LOD Bias: This signed value is added to the LOD when the LOD is determined at a
textured pixel. This 2’s complement, fixed-point value has 2 integer bits and 4 fractional bits
and is signed extended before being added to the LOD. The valid data value range is from -4.0
to 3.9375. (S2.4) The default value is 0.
257
®
R
13.11 GFXRENDERSTATE_MAP_PALETTE_LOAD
The Texture Palette is loaded using the following instruction. All 256 entries of the texture palette must be loaded every
time this command packet is sent. In other words, even if only one entry of the texture palette must be updated, all 256
entries must be loaded and sent using this command packet.
Coordinate Sets
Map Information
Texels
Blending Stages
DWord
Bit
0
31:29
28:24
3DStateMWNP (non-pipelined): 1Dh
23:16
Opcode: 82h
15:0
DWORD_LENGTH: 255
31:16
Reserved: 00h
15:0
16-bit Color[0]: Format: RGB 565/1555/4444, YCrCb 565 or AI 88
1
2-255
256
258
Description
…
31:16
Reserved: 00h
15:0
16-bit Color[255]: Format: RGB 565/1555/4444, YCrCb 565 or AI 88
®
R
13.12 GFXRENDERSTATE_MAP_COLOR_BLEND_STAGES
The rendering engine supports three map color blend stages for the red, green, and blue channels. Any of these stages
may perform an operation utilizing up to two texels from the mapping engine, the iterated face color, the iterated alpha
and/or a constant color.
Coordinate Sets
Map Information
Texels
Blending Stages
The following equations are supported by each blending stage.
Blend Equation
Description
Colorout = Arg1
Select Arg1
Colorout = Arg2
Select Arg2
Colorout = Arg1 * Arg2
Modulate
Colorout = Arg1 * Arg2 * 2
Modulate and Multiply by 2
Colorout = Arg1 * Arg2 * 4
Modulate and Multiply by 4
Colorout = Arg1 + Arg2
Add
Colorout = Arg1 + Arg2 - 0.5
Add Signed (Excess 128)
Colorout = αIterated * Arg1 + (1 - αIterated) * Arg2
Linearly Blend using Iterated Alpha
Colorout = αFactor * Arg1 + (1 - αFactor) * Arg2
Linearly Blend using Alpha Factor
Colorout = αTexel0 * Arg1 + (1 - αTexel0) * Arg2
Linearly Blend using Texel0’s Alpha
Colorout = αTexel1 * Arg1 + (1 - αTexel1) * Arg2
Linearly Blend using Texel1’s Alpha
Colorout = ColorTexel0 * Arg1 + (1 - ColorTexel0) * Arg2
Linearly Blend using Texel0’s Color
Colorout = ColorTexel1 * Arg1 + (1 - ColorTexel1) * Arg2
Linearly Blend using Texel1’s Color
The settings for these stages are specified in the following instruction. When a stage for color operation is enabled, the
corresponding alpha stage also will be enabled. The arguments and operation fields of the corresponding alpha stage
must be programmed accordingly.
259
®
DWord
0
Bit
R
Description
31:29
28:24
3DState24: 00h
23:22
Reserved: 0h
21:20
Update Blending Stage Index: The valid range is 0-2.
19
Current/Accumulator Select Mask (0: do not update; 1: update)
18
Write result to Current Register or Accumulator Select
0 - Current register (default)
1- Accumulator
17
Color Arg1 Mask (0: do not update; 1: update)
16:14
Color Arg1: Valid values are:
0h – All ones
1h – Color factor
2h – Accumulator color (illegal setting for stage 0)
3h – Iterated color
4h – Specular color
5h – Current color
6h – Texel0 color
7h – Texel1 color
13
Replicate Arg1 Alpha to Color Channels:
0 – Do not replicate alpha to color channels.
1 – Replicate alpha to color channels.
12
Invert Color Arg1:
0 – Do not invert argument.
1 – Invert argument.
0
11
Color Arg2 Mask (0: do not update; 1: update)
10:8
Color Arg2: Valid values are:
0h – All ones
1h – Color factor
2h – Accumulator color (illegal setting for stage 0)
3h – Iterated color
4h – Specular color
5h – Current color
6h – Texel0 color
7h – Texel1 color
7
Replicate Arg2 Alpha to Color Channels:
0 – Do not replicate alpha to color channels.
1 – Replicate alpha to color channels.
0
6
Invert Color Arg2:
260
®
R
DWord
Bit
Description
5
Color Operation Mask (0: do not update; 1: update)
4:0
Color Operation: Valid values are:
00h – Disable this stage and all higher-numbered stages.
(This bit disables/enables both the color and the corresponding alpha stage.)
01h – Select Arg1
02h – Select Arg2
03h – Modulate
04h – Modulate and multiply by 2.
06h – Add
07h – Add signed
08h – Linearly blend using the iterated alpha as the blend term.
09h – Reserved
0ah – Linearly blend using the alpha factor as the blend term.
0bh-0fh – Reserved
10h – Linearly blend using texel0’s alpha as the blend term.
12h – Linearly blend using texel0’s color as the blend term.
13h – Linearly blend using texel1’s color as the blend term.
14h – Subtract (Arg1 - Arg2)
15h-1fh – Reserved
261
®
R
13.13 GFXRENDERSTATE_MAP_ALPHA_BLEND_STAGES
The rendering engine supports three map color blend stages for the alpha channel. Any of these stages may perform an
operation utilizing the alpha channel from up to two texels from the mapping engine, the iterated alpha and/or a constant
color.
Coordinate Sets
Map Information
Texels
Blending Stages
The following equations are supported by each blending stage.
Blend Equation
Description
Alphaout = Arg1
Select Arg1
Alphaout = Arg2
Select Arg2
Alphaout = Arg1 * Arg2
Modulate
Alphaout = Arg1 * Arg2 * 2
Modulate and multiply by 2.
Alphaout = Arg1 * Arg2 * 4
Modulate and multiply by 4.
Alphaout = Arg1 + Arg2
Add
Alphaout = Arg1 + Arg2 - 0.5
Add signed (excess 128).
Alphaout = αIterated * Arg1 + (1 - αIterated) * Arg2
Linearly blend using iterated alpha.
Alphaout = αFactor * Arg1 + (1 - αFactor) * Arg2
Linearly blend using alpha factor.
Alphaout = αTexel0 * Arg1 + (1 - αTexel0) * Arg2
Linearly blend using texel0’s alpha.
Alphaout = αTexel1 * Arg1 + (1 - αTexel1) * Arg2
Linearly blend using texel1’s alpha.
262
®
R
The settings for these stages are specified in the following instruction.
DWord
Bit
Description
0
31:29
28:24
3DState24: 01h
23:22
Reserved: 0h
21:20
Update Blending Stage Index: The valid range is 0-2.
19
Reserved: 0h
18
Alpha Arg1 Mask (0: do not update; 1: update)
17:15
Alpha Arg1: Valid values are:
0h – Reserved
1h – Alpha factor
2h – Reserved
3h – Iterated alpha
4h – Reserved
5h – Current alpha
6h – Texel0 alpha
7h – Texel1 alpha
14
Reserved: 0h
13
Invert Alpha Arg1:
12
Alpha Arg2 Mask (0: do not update; 1: update)
11
Reserved: 0h
10:8
Alpha Arg2: Valid values are:
0h – Reserved
1h – Alpha factor
2h – Reserved
3h – Iterated alpha
4h – Reserved
5h – Current alpha
6h – Texel0 alpha
7h – Texel1 alpha
7
Reserved: 0h
6
Invert Alpha Arg2:
0
5
Alpha Operation Mask (0: do not update; 1: update)
263
®
DWord
Bit
0
4:0
R
Description
Alpha Operation: Valid values are:
00h – Reserved
(Enable/disable of stage is programmed in the color operation field.)
01h – Select Arg1
02h – Select Arg2
03h – Modulate
06h – Add
07h – Add signed
08h – Linearly blend using the iterated alpha as the blend term.
09h – Reserved
0ah – Linearly blend using the alpha factor as the blend term.
0bh-0fh – Reserved
12h-1fh – Reserved
13.14 GFXRENDERSTATE_COLOR_FACTOR
This instruction specifies a constant color factor that can be used in the blending stages.
DWord
0
1
264
Bit
Description
31:29
28:24
3DStateMW: 1Dh
23:16
Instruction: 01h
15:0
DWORD_LENGTH: 0000h
31:0
Color Factor: This is a 32-bit value in the ARGB 8888 format. The top 8 bits are used by the
alpha blending stage, and the lower 24 bits are used by the color blending stage.
®
R
13.15 GFXRENDERSTATE_COLOR_CHROMA_KEY
ColorKey and chromakey are terms used to describe two methods of removing a specific color or range of colors from a
map that is applied to a primitive.
When a color palette is used with indices to indicate a color in the palette, the indices can be compared against a state
variable “ColorKey Index Value.” If a match occurs and ColorKey is enabled, then an action will be taken to remove this
value’s contribution to the resulting pixel color.
The chromakey mode refers to testing the RGB or YUV components to see if they fall between a high
(Chroma_High_Value) and low (Chroma_Low_Value) state variable values. If the color of a texel contribution is in this
range and chromakey is enabled, then an action will be taken to remove this contribution to the resulting pixel color.
When the color value of the nearest neighbor texel falls within the range of the chromakey values or the palletized texel
index matches the chromakey value, the pixel can be programmed, by means of the killpixel state variable, to either be or
not be written back to the frame buffer. The texture alpha value is always set to zero when this occurs.
When multiple texture is enabled, only the color of texel 0 is used for chromakey or ColorKey. The color of texel 1 is
ignored in these comparisons and does not contribute to the result.
DWord
Bit
0
31:29
28:24
3DStateMW: 1Dh
23:16
Instruction: 2h
15:0
DWORD_LENGTH: 1
31:29
Reserved: 00h
1
Description
28
Kill Pixel Write Mask (0: do not update; 1: update)
27
Kill Pixel Write When Match on Key (0: do not kill pixel; 1: kill pixel)
Enabling the Kill Pixel Write state prevents the pixel from being written into the destination
buffer, independently of any alpha test. The alpha value is always set to zero for keyed-out
pixels.
2
26
Color Index Value Mask (0: do not update; 1: update)
25
Chromakey Low Value Mask (0: do not update; 1: update)
24
Chromakey High Value Mask (0: do not update; 1: update)
23:0
Chromakey Low Value: This is a 24-bit value in RGB 888 format. Only the upper 5/6/5 bits
of the RGB 888 color are compared with the texel colors.
31:24
Color Index Value: This is an 8-bit color index. The default value is 0.
23:0
Chromakey High Value: This is a 24-bit value in RGB 888 format. Only the upper 5/6/5
bits of the RGB 888 color are compared with the texel colors.
265
®
R
13.16 GFXRENDERSTATE_SRC_DST_BLEND_MONO
The 3D section of the GC supports the alpha blend modes defined in the following state variables, except where noted as
“Not Implemented.”
RENDERSTATE_SRC_DST_BLEND
RENDERSTATE_BLEND_ENABLE
RENDERSTATE_BLEND_COLOR (texture compositing)
Additional Blending Modes
The color calculator super-unit of the GC supports four blending operations. The blending operations use a constant
color and constant alpha. The blend factors for the source and destination colors are controlled by separate state variables
and are defined as follows. The RGBA values of the source and destination are indicated with the subscripts s and d,
respectively, while the RGBA constant values are indicated by the subscript c.
266
R
®
The Zero blend factor is:
Rout = 0
Gout = 0
Bout = 0
Aout = 0
The One blend factor is:
Rout = Rin
Gout = Gin
Bout = Bin
Aout = Ain
The Src_Color blend factor is:
Rout = Rs * Rin
Gout = Gs * Gin
Bout = Bs * Bin
Aout = As * Ain
The Inv_Src_Color blend factor is:
Rout = (1 - Rs) * Rin
Gout = (1 - Gs) * Gin
Bout = (1 - Bs) * Bin
Aout = (1 - As) * Ain
The Src_Alpha blend factor is:
Rout = As * Rin
Gout = As * Gin
Bout = As * Bin
Aout = As * Ain
The Inv_Src_Alpha blend factor is:
Rout = (1 - As) * Rin
Gout =(1 - As) * Gin
Bout = (1 - As) * Bin
The Dst_Color blend factor is:
Rout = Rd * Rin
Gout = Gd * Gin
Bout = Bd * Bin
Aout = Ad * Ain
267
®
The Inv_Dst_Color blend factor is:
Rout = (1 - Rd) * Rin
Gout = (1 - Gd) * Gin
Bout = (1 - Bd) * Bin
Aout = (1 - Ad) * Ain
The Both_Src_Alpha blend factor is:
Source:
Rout = As * Rin
Gout = As * Gin
Bout = As * Bin
Aout = As * Ain
Destination:
Gout = (1 - As) * Gin
The Both_Inv_Src_Alpha blend factor is:
Source:
Gout = (1 - As) * Gin
Destination:
Rout = As * Rin
Gout = As * Gin
Bout = As * Bin
Aout = As * Ain
268
R
®
R
The blending factors are applied to the source and destination RGBA values and computed using one of the following
blending equations. The equation is selected by a state variable.
Add: Cs * S + Cd * D
where Cs is the source RGBA, Cd is the destination RGBA, S is the source blending factor, and D is the destination
blending factor.
Dword
Bit
Description
0
31:29
28:24
3DState24: 08h
23:14
Reserved: 00h
13
Color Monochrome Enable Statemask (1: update; 0: do not update)
12
Color Monochrome Enable: (1: enable; 0 = disable (default))
11
Source Blend State Mask (1: update; 0: do not update)
10:6
Source Blend State: (see the following table)
5
Destination Blend State Mask: (1: update; 0: do not update)
4:0
Opcode
Destination Blend State: (see the following table)
Source / Destination Blend State
00h
Reserved (default)
01h
Zero
02h
One
03h
Src_Color
04h
Inv_Src_Color
05h
Src_Alpha
06h
Inv_Src_Alpha
07h
Reserved
08h
Reserved
09h
Dst_Color
0ah
Inv_Dst_Color
0bh
Reserved
0ch
Both_Src_Alpha
0dh
Both_Inv_Src_Alpha
In the case of the monochrome state variables, the rasterizer will shade primitives by interpolating the blue color and
using the result for the red, green, and blue color components for specular and diffuse colors.
269
®
R
13.17 GFXRENDERSTATE_Z_BIAS_ALPHA_FUNC_REF
DWord
Bit
Description
0
31:29
Client: 03h – Render Processor
28:24
3DState24NP (non-pipelined): 14h
23
Reserved: 0
22
Z Bias State Mask (1: update; 0: do not update)
21:14
13
12:9
8
Z Bias: This signed value is added to the source Z value prior to the Z compare function.
This 2’s complement fixed-point value has 7 integer bits and 0 fractional bits and is sign
extended before being added to the destination Z value. The valid data value range is from
-128 to 127. (S7) The default value is 0. The Z values written back to the Z-buffer include
this Z bias value.
Alpha Function State Mask: (1: update; 0: do not update)
Alpha Function: Valid values are:
00h – Reserved
01h - Never (Never pass.)
02h - Less (Pass if the source alpha is less than the alpha reference.)
03h - Equal (Pass if the source alpha is equal to the source alpha.)
04h – Lequal (Pass if the source alpha is less than or equal to the source alpha.)
05h – Greater (Pass if the source alpha is greater than the source alpha.)
06h – Notequal (Pass if the source alpha is note equal to the source alpha.)
07h – Gequal (Pass if the source alpha is greater than or equal to the source alpha.)
08h – Always (Always pass.)
Alpha Reference State Mask: (1: update; 0: do not update)
7:3
Alpha Reference value for color calculation equation(s): This is bits 7-3 of an alpha
reference value. This value specifies a reference alpha value against which pixels are
tested when alpha-testing is enabled. The default value is 0. (unsigned int)
2:0
Reserved: 000
The Z Bias state variable specifies the value used to offset the source z value before performing the z compare test,
which is referred to as backend Z bias. This is used for coplanar polygon priority. If two coplanar polygons are to be
rendered, because of the inherent precision differences induced by unique x, y, and z values, there is no guarantee as to
which polygon will be closer or farther. By using the Z bias state variable, it is possible to offset the source z value
(compare value) before comparing with the destination z value. The Z bias value is added at the LSBs of a 16-bit z
representation. Using the Z bias state variable affects the source z value, and the result of the z-bias addition is written to
the z buffer.
270
®
R
13.18 GFXRENDERSTATE_LINE_WIDTH_CULL_SHADE_MODE
The provoking vertex refers to the vertex that selected the flat-shaded color for the primitive. In the OpenGL
specification, the third/second (triangle/line) vertex should used. In D3D specification, the first vertex should be used.
There is an additional mode that allows the selection of the common vertex for the triangle fan primitive and the
third/second (triangle/line) vertex for other primitives.
The GC supports both OpenGL and D3D primitive rasterization rules that determine whether a pixel is filled by the
triangle or line. Both D3D and OpenGL use a top-left filling convention for filling geometry. The center of the pixel is
the point at which decisions are made. If the center is inside a triangle, the pixel is part of the triangle. If an edge of a
primitive intersects the center of a pixel, the following rules will be used to determine if that pixel is part of the primitive.
This rule is independent of the orientation of the primitive.
OpenGL Rasterization Rules:
OpenGL specifies that a pixel is included in a primitive if its pixel center lies inside the primitive. Special treatment
is required in the case of a pixel whose center lies on a polygon edge. In such a case, if two primitives share a
common edge on which a pixel center lies, then exactly one of the primitives can include that pixel.
The center of a pixel is defined as X.5,Y.5, where X and Y are any screen coordinates.
D3D Rasterization Rules:
For all primitives, a pixel is included in a primitive if the edge that intersects the pixel center is the left edge of that
primitive. If the edge that intersects the pixel center is exactly horizontal, the pixel is included in the primitive if the
intersecting edge is the top edge of that primitive.
The center of a pixel is defined as X.0,Y.0, where X and Y are any screen coordinates
Please note that this instruction packet affects the following selections:
D3D and OpenGL strip and fan notations selection
D3D and OpenGL vertex selection on flat shading color
D3D and OpenGL rasterization rule selection
In the flat-shaded mode, the value of the first vertex in the primitive is used to determine the value of the face. In the
Gouraud-shaded mode, the value of the face is determined by a linear interpolation between all of the primitive’s
vertices.
DWord
0
Bit
Description
31:29
28:24
3DState24: 02h
0
23
Reserved
0
22:21
Reserved
20
Z Function State Mask: (1: update; 0: do not update)
19
Reserved
271
®
DWord
0
272
Bit
R
Description
18:16
Z Function: Valid values are:
00h - Always (Always pass.) (default)
01h - Never (Never pass.)
02h - Less (Pass if the source Z is less than the destination Z.)
03h - Equal (Pass if the source Z is equal to the destination Z.)
04h - Lequal (Pass if the source Z is less than or equal to the destination Z.)
05h - Greater (Pass if the source Z is greater than the destination Z.)
06h - Notequal (Pass if the source Z is note equal to the destination Z.)
07h - Gequal (Pass if the source Z is greater than or equal to the destination Z.)
15
Line Width State Mask: (1: update; 0: do not update)
14:12
Line Width: This value specifies the width of lines, in pixels. The variable is represented as an
unsigned value with 2 integer bits and 1 fractional bit. The valid value range is from 0.0 to 3.5,
in 0.5 increments. The default value is 0.
11
Alpha Shade Mode State Mask (1: update; 0: do not update)
10
Alpha Shade Mode (1: flat, 0: Gouraud (default))
9
Fog Shade Mode State Mask (1: update; 0: do not update)
8
Fog Shade Mode (1: flat, 0: Gouraud (default))
7
Specular Shade Mode State Mask (1: update; 0: do not update)
6
Specular Shade Mode (1: flat, 0: Gouraud (default))
5
Color Shade Mode State Mask (1: update; 0: do not update)
4
Color Shade Mode (1: flat, 0: Gouraud (default))
3
Cull Mode State Mask (1: update; 0: do not update)
2:0
Cull Mode: To support the culling of back-facing triangles
000 – Reserved
001 – Cull no triangle (no back face removal).
010 – Cull CW triangles (allow CCW triangles).
011 – Cull CCW triangles (allow CW triangles).
100 – Cull both (for performance experiments).
101 to 111 – Reserved
®
R
13.19 GFXRENDERSTATE_BOOLEAN_ENA_1
DWord
Bit
Description
0
31:29
Client: 03h – Render Engine
28:24
3DState24: 03h
23:18
Reserved: 00h
17
Alpha Setup Enable Mask (1: update; 0: do not update)
16
Alpha Setup Enable (0: disable (default), 1: enable)
This SV enables alpha gradient calculations in the Setup Unit (independently of other alpha
SVs).
15:8
Reserved: 00h
7
Fog Enable State Mask (1: update; 0: do not update)
6
Fog Enable (1: enable, 0: disable (default))
5
Alpha Enable State Mask (1: update; 0: do not update)
4
Alpha Test Enable (1: enable, 0: disable (default))
3
Blend Enable State Mask (1: update; 0: do not update)
2
Blend Enable (1: enable, 0: disable (default))
1
Z Enable State Mask (1: update; 0: do not update)
0
Z Enable (1: enable, 0: disable (default))
273
®
R
13.20 GFXRENDERSTATE_BOOLEAN_ENA_2
DWord
Bit
Description
0
31:29
28:24
3DState24: 04h
23:18
Reserved: 00h
17
Mapping Cache Enable Mask (1: update; 0: do not update)
16
Mapping Cache Enable (1: enable, 0: disable (default))
15
Alpha Dither Enable Mask (1: update; 0: do not update)
14
Alpha Dither Enable (1: enable, 0: disable (default))
13
Fog Dither Enable Mask (1: update; 0: do not update)
12
Fog Dither Enable (1: enable, 0: disable (default))
11
Specular Dither Enable Mask (1: update; 0: do not update)
10
Specular Dither Enable (1: enable, 0: disable (default))
9
Color Dither Enable Mask (1: update; 0: do not update)
8
Color Dither Enable (1: enable, 0: disable (default))
7:4
Reserved: 0h
3
Frame Buffer Write Enable Mask (1: update; 0: do not update)
2
Frame Buffer Write Enable (1: enable, 0: disable (default))
1
Z Buffer Write Enable Mask (1: update; 0: do not update)
0
Z Buffer Write Enable (1: enable, 0: disable (default))
The frame buffer write enable feature can be used to reliably render coplanar (layered) polygons. An example of this
type of layering would be to render a window on a wall. With a 16-bit z buffer, there is insufficient precision to render
this type of scene with consistent priority. Some pixels of the window would be intermixed with some pixels of the wall,
due to precision inaccuracies.
274
®
R
13.21 GFXRENDERSTATE_FOG_COLOR
The GFXRENDERSTATE_FOG_COLOR state instruction format is as follows:
DWord
Bit
Description
0
31:29
28:24
3DState24NP (non-pipelined): 15h
23:19
Fog Color Red: Bits 7 - 3 of the red fog color component. The default value is 0. (unsigned int)
18:16
Reserved: 0h
15:10
Fog Color Green: Bits 7 - 2 of the green fog color component. The default value is 0. (unsigned int)
9:8
Reserved: 0h
7:3
Fog Color Blue: Bits 7 - 3 of the blue fog color component. The default value is 0. (unsigned int)
2:0
Reserved: 0h
275
®
R
13.22 GFXRENDERSTATE_DRAWING_RECTANGLE_INFO
The values programmed in this packet are in screen coordinates. In the full-screen mode, the screen pitch and height
minus one must be programmed into this packet. For example, for the screen size 1280×1024, the programmed values
should be Xmin = 0, Ymin = 0, Xmax = 1279, and Ymax = 1023. When rendering rectangle primitives, the height and
width of the drawing rectangle must be greater than 1. For example, (Xmax - Xmin) > 1 and (Ymax - Ymin) > 1.
The drawing rectangle coordinates must be clipped to the screen boundary by software before being sent down to the
hardware.
DWord
Bits
0
31:29
28:24
23:16
Opcode: 80h
15:0
DWORD_LENGTH: 3
1
31
Description
Drawing/scissor rectangle clipping (for validation purpose only)
0: Enable
1: Disable
2
3
4
276
30-28
Reserved: 0h
27-26
A 2-bit value of x bias for dithering compensation
25-24
A 2-bit value of y bias for dithering compensation
23-0
Reserved: 000000h
31:16
Clipped DR Ymin: Y coordinate of upper-left corner of drawing rectangle after being clipped
to screen boundary. Only the lower 10 bits in U10 format are used by hardware. This value is
a pixel-aligned, unsigned integer, and it is referenced to the upper-left corner of the 2D buffer
defined in the dest buffer info packet.
15:0
Clipped DR Xmin: X coordinate of upper-left corner rectangle after being clipped to screen
boundary. Only the lower 11 bits in U11 format are used by hardware. This value is a pixelaligned, unsigned integer, and it is referenced to the upper-left corner of the 2D buffer defined
in the dest buffer info packet.
31:16
Clipped DR Ymax: Y coordinate of lower-right corner rectangle after being clipped to screen
15:0
Clipped DR Xmax: X coordinate of lower-right corner rectangle after being clipped to screen
31:27
Reserved
26:16
DR origin Y: Origin of the Y coordinate of the drawing rectangle before being clipped to
screen boundary. The value is S10 in 2’s complement format. This value is pixel-aligned,
referenced to the upper-left corner of the 2D buffer defined in the dest buffer packet.
15:12
Reserved
®
R
DWord
Bits
11:0
Description
DR origin X: Origin of the X coordinate of the drawing rectangle before being clipped to
screen boundary. The value is S11 in 2’s complement format. This value is pixel-aligned,
referenced to the upper-left corner of the 2D buffer defined in the dest buffer packet.
13.23 GFXRENDERSTATE_SCISSOR_ENABLE
Only inclusive mode scissoring is supported.
Span-Aligned Reference
Point Defined by the Back
Buffer Info Packet
“Inclusive”
Clipping
Pixel-Aligned
Clipping Rectangle
Pixel-Aligned
Drawing Window
Drawn Region
277
®
DWord
Bit
0
31:29
28:24
3DState16NP (non-pipelined): 1Ch
23:19
Opcode: 10h
18:17
Reserved: 0h
16
15:2
R
Description
Scissor Rectangle ID: This number must be 0 for this stepping, since only one clipping rectangle
is allowed. (More clipping rectangles can be allowed in future expansion.)
Reserved: 0000h
1
Scissor Rectangle Enable Mask (1: update; 0: do not update)
0
Scissor Rectangle Enable (0 = disable (default), 1 = enable)
13.24 GFXRENDERSTATE_SCISSOR_RECTANGLE_INFO
The coordinate in this instruction packet is relative to the upper-left corner (i.e., the origin) of the clipping rectangle
defined in the Clipping_Rectangle_Info packet.
DWord
Bits
0
31:29
28:24
23:16
Opcode: 81h
15:0
DWORD_LENGTH: 1
31:16
Ymin: Y coordinate of upper-left corner. Only the lower 10 bits are used by hardware. This
value is a pixel-aligned, unsigned integer, referenced to the upper-left corner of the drawing
rectangle.
15:0
Xmin: X coordinate of upper-left corner. Only the lower 11 bits are used by hardware. This
value is a pixel aligned, unsigned integer, referenced to the upper-left corner of the drawing
rectangle.
31:16
Ymax: Y coordinate of lower-right corner. Only the lower 10 bits are used by hardware. This
rectangle.
15:0
Xmax: X coordinate of lower-right corner. Only the lower 11 bits are used by hardware. This
rectangle.
1
2
278
Description
®
R
13.25 Stipple Pattern
The stipple pattern is a 4×4-bit memory that serves as pixel write mask. When stippling is enabled, the frame buffer will
only be updated with pixels that have ones in their associated stipple pattern memory locations. The stipple pattern
memory maps to a 4×4-pixel grid. The stipple pattern repeats for x and y coordinates on every span (4×4 pixels). The
stipple pattern is mostly for testing purposes. This is a non-pipelined state variable. The stipple pattern pixel enable
function can be represented as follows:
Word
Bits
0
31:29
28:23
22:16
Opcode: 83h
15:0
DWORD_LENGTH: 0
31:17
Reserved
1
16
15:0
Description
Enable: 1 = enable, 0 = Disable
Stipple pattern
13.26 GFXRENDERSTATE_ANTI_ALIASING
The antialiasing packet defines the state variables for antialiasing enable, bounding-box expansion, line antialiasing
region width, polygon antialiasing region width, and edge flag enable.
The antialiasing enable state variable turns on antialiasing of polygons and lines. The bounding box enclosing the
triangle or line is expanded in all directions by the bounding box expansion value (in pixels). The line antialiasing region
width state variable allows for the specification of the width over which to blend for the antialiasing of lines. Similarly,
the polygon antialiasing region width state variable allows for the specification of the width over which to blend for the
antialiasing of polygon edges. The edge flag enable state variable enables all the edge flags of a triangle/line overriding
the edge flags specified in the vertex format.
279
®
DWord
Bits
0
31:29
28:24
3DState24: 06h
23:14
Reserved: 000h
Description
13
Edge Flag Enable Mask: 1 = update; 0 = do not update
12
Edge Flag Enable: 1 = enable; 0 = disable
11
Polygon Antialiasing Region Width Enable Mask: 1 = update, 0 = do not update
10:9
8
7:6
5
4:2
280
R
Polygon Antialiasing Region Width: Valid values are:
0h - 0.5 pixels
1h - 1.0 pixels
2h - 2.0 pixels
3h - 4.0 pixels
Line Antialiasing Region Width Enable Mask: 1 = update, 0 = do not update
Line Antialiasing Region Width: Valid values are:
0h - 0.5 pixels
1h - 1.0 pixels
2h - 2.0 pixels
3h - 4.0 pixels
Bounding Box Expansion Mask: 1 = update, 0 = do not update
Bounding Box Expansion: U3.0 format (unsigned) with valid values from 0 to 7 pixels
1
Antialiasing Enable Mask: 1 = update, 0 = do not update
0
Antialiasing Enable: 1 = enable, 0 = disable
®
R
13.27 GFXRENDERSTATE_PROVOKING_VTX_PIXELIZATION_RULE
The provoking vertex state variables provide the flexibility of selecting the flat-shaded vertex for the first triangle/line of
a primitive packet. The subsequent flat-shaded vertices are in incremental order, except in the case of triangle fans with
common flat-shaded vertices. Separate state variables are provided for triangles and lines. The following tables show the
sequence of flat-shaded vertices for triangles and lines:
Primitive Type
Triangle/Rectangle List
“
“
Triangle
Strip/List Flat
Shading SV
00
01
10
Triangle Fan Flat
Shading SV
Line Strip/List
Flat Shading SV
Flat-Shaded Vertex
Sequence
XX
XX
XX
XX
XX
XX
0, 3, 6, 9, 12...
1, 4, 7, 10, 13...
2, 5, 8, 11, 14...
Triangle Strip/Triangle
Strip Reverse Winding
Order
“
“
00
XX
XX
0, 1, 2, 3, 4…
01
10
XX
XX
XX
XX
1, 2, 3, 4, 5…
2, 3, 4, 5, 6…
Triangle Fan
“
“
XX
XX
XX
00
01
10
XX
XX
XX
0, 0,0,0,0,0…
1, 2, 3, 4, 5…
2, 3, 4, 5, 6…
Polygon (Triangle Fan
With Common FlatShaded Vertex)
XX
XX
XX
0,0,0,0,0,0…
Line List
“
XX
XX
XX
XX
X0
X1
0, 2, 4, 6, 8…
1, 3, 5, 7, 9…
Line Strip
“
XX
XX
XX
XX
X0
X1
0, 1, 2, 3, 4…
1, 2, 3, 4, 5…
The Pixelization Rule state-variable sets up the hardware to follow correct pixelization rules. It should be set in the
Render Mode (i.e., for the GFXPrimitive packet). Small-Triangle Filter Enable enables filtering of small triangles.
(Small Triangles are defined as triangles that do not light any pixels.) It should be set only in Render Mode when the
Pixelization Rule is enabled, and the horizontal & vertical bias values are equal and are programmed to either 1000 or
0000. If other bias values are used, then the results may be unpredictable. (The horizontal & vertical bias values are
programmed in the GFXRENDERSTATE_DEST_BUFFER_INFO packet). The following table summarizes the
acceptable values for these SVs:
281
®
Packet
GFXPRIMITIVE
GFXPRIMITIVE
GFXPRIMITIVE
GFXPRIMITIVE
Horizontal Bias
0000
1000
XXXX
XXXX
R
Vertical Bias
0000
1000
XXXX
XXXX
DWord
Bits
0
31:29
28:24
3DState24: 07h
23:13
Reserved: 000h
Small-Triangle Filter
1
1
0
0
Description
12
Small-Triangle Filter Enable Mask (1: update, 0: do not update)
11
Small-Triangle Filter Enable: If set, then small triangles (i.e., triangles that do not cover any
pixels and are either vertical or horizontal) will be filtered out. Note that this should be
enabled only in the GFXPrimitive mode, when the pixelization rule is enabled and the
“horizontal & vertical bias” values (in the Dest Buffer Packet) are both the same and
programmed to 1000 or 0000.
10
Pixelization Rule Mask
1: update 0: do not update
9
Pixelization Rule: This bit should be set in the GFXPRIMITIVE mode for following the correct
pixelization rules.
8
Line Strip/List Flat-Shaded Vertex Select Mask (1 = update, 0 = do not update)
7:6
5
4:3
2
1:0
282
Pixelization Rule
1
1
1
0
Line Strip/List Flat-Shaded Vertex Select:
0h - Vertex 0
1h - Vertex 1
2h – Reserved
3h – Reserved
Triangle Fan Flat-Shaded Vertex Select Enable Mask (1 = update, 0 = do not update)
Triangle Fan Flat-Shaded Vertex Select:
0h - Vertex 0
1h - Vertex 1
2h - Vertex 2
3h - Reserved
Triangle Strip/List Flat-Shaded Vertex Select Enable Mask (1 = update, 0 = do not update)
Triangle Strip/List Flat-Shaded Vertex Select:
0h - Vertex 0
1h - Vertex 1
2h - Vertex 2
3h - Reserved
®
R
13.28 GFXRENDERSTATE_DEST_BUFFER_VARIABLES
The GFXRENDERSTATE_DEST_BUFFER_VARIABLES instruction is used to specify information about the
destination buffer. This information is used to initialize the rendering hardware parameters. The destination buffer
contains the pixel color data for the scene being rasterized by the 3D rendering engine. This is a non-pipelined state
variable.
DWord
Bits
0
31:29
28:24
23:16
Opcode: 85h
15:0
DWORD Length: 0
31:24
Reserved: 00h
23:20
Destination Origin Horizontal Bias: This is an unsigned value (0.4) that is used to bias the origin of
the X values associated with the vertices of a primitive. The unbiased origin is located in the upperleft corner of a square pixel, with the center located at 0.5, 0.5. A bias value of ½ (8h) would move
the origin toward the right, such that an X value of 0.0 would specify the center of a pixel.
19:16
Destination Origin Vertical Bias: This is an unsigned value (0.4) that is used to bias the origin of the
Y values associated with the vertices of a primitive. The unbiased origin is located in the upper-left
corner of a square pixel, with the center located at 0.5, 0.5. A bias value of ½ (8h) wood move the
origin toward the bottom, such that an Y value of 0.0 would specify the center of a pixel.
15:11
Reserved: 00h
13:12
4:2:2 Channel Write Select: Select which channel(s) are written to the destination buffer
when the surface format is 4:2:2, using byte masks.
1
Description
00 – Write all channels (Y,Cr, and Cb)
01 – Write only the Y channel
10 – Write only the Cr channel
11 – Write only the Cb channel
11
10:8
Reserved: 0h
Destination Buffer Format:
0h – Any 8-bit surface
1h – RGB 555
2h – RGB 565
4h – 4:2:2 YCrCb (Y swap format in memory)
5h – 4:2:2 YcrCb (Normal format in memory)
6h – 4:2:2 YcrCb (UV swap format in memory)
7h – 4:2:2 YcrCb (UV/Y swap format in memory)
7:2
1
Reserved: 00h
Vertical Line Stride: The number of lines to skip between logically adjacent lines
0 – None
1 – One
283
®
0
Vertical Line Stride Offset: The number of lines to add as an initial offset, when the vertical line
stride is 1
0 – None
1 – One
284
R
R
®
14. Clock Control Registers
The Intel 82810 Chipset has three PLLs to generate all the clocks. The clock control registers are accessed by writing to
the memory-mapped address offset. The host PLL generates the host clock, whose frequency is controlled by an external
strap. In addition, the host PLL generates the system and local memory core clock and graphics core clock. The hub PLL
generates the clock for the hublink unit. The display PLL generates the display or LCD clock.
The display clock can be controlled by three blocks of registers: DCLK0, DCLK1, and DCLK2. Each display clock has
its own Display Clock i Divisor registers for M, N, and a byte of Display & LCD Clock Divisor Select Register,
within which are P (Divisor) values, and can be programmed independently. DCLK0 and DCLK1 normally are
programmed to 25.175 MHz and 28.322 MHz, respectively (VGA-compatible clocks). DCLK2 is used for non-VGA
modes.
The Display Clock i Divisor register and the appropriate byte of Display & LCD Clock Divisor Select Register are
programmed with the loop parameters to be loaded into the clock synthesizer. The MSR[3:2] register is used to select
among DCLK0(default), DCLK1 and DCLK2. Writing to LCD/TV Out Control [31] = 1 and [0] = 1 selects the LCD
clock. MSR[3:2] are ignored when this condition is true.
The data written to these registers are calculated based on the reference frequency, the desired output frequency, and the
characteristic VCO constraints, as described in the Functional Description. From the calculation, the M, N, and P values
are obtained.
14.1 Programming Notes
The three blocks of registers (dclk0, dclk1, dclk2) are used to program up to three unique frequencies for the display
clock. The blocks of registers can be programmed independently of each other. However, only one can be selected at any
point in time to control the DPLL. The MSR register (bits 3:2) is used to determine which of the DCLK0,1,2 register
groups will control the DPLL. Writing to MSR register bits 3:2 also transfers the Display Clock Divisor and Display &
LCD Clock Divisor Select Register contents to the VCO register file.
285
®
Write the Display Clock 0 Divisor Register with the M-REG value and N-REG value.
Write MSR register bits 3:2 = 00 to select DCLK0. (NOTE: This is the default value.)
Write the Display Clock 1 Divisor Register with the M-REG value and N-REG value.
Write MSR register bits 3:2 = 01 to select DCLK1.
Write MSR register bit 3 = 1 to select DCLK2.
Example Programming Sequence (LCD CLK)
Write the LCD Clock Divisor Register with the M-REG value and N-REG value.
Write the LCD byte of the Display & LCD Clock Divisor Select Register with the P-REG value.
Write the LCD / TV Out Control[31] = 1 and [0] = 1. MSR[3:2] are ignored when this condition is TRUE.
286
R
®
R
14.2 DCLK_0D — Display Clock 0 Divisor Register
Address offset:
06000h–06003h
Default:
00030013h
Attributes:
R/W
Size:
32 bits
The Display Clock 0 Divisor Register and Display & LCD Clock Divisor Select Register are programmed with the
The data written to this register are calculated based on the reference frequency, the desired output frequency, and the
characteristic VCO constraints.
31
26
Reserved
Bit
25
VCO 0 N-divisor
16
15
10
Reserved
9
0
VCO 0 M-divisor
Description
31-26
Reserved
25-16
VCO 0 N-Divisor. N-divisor value calculated for the desired output frequency. (Default = 03h)
15-10
Reserved
9-0
VCO 0 M-Divisor. M-divisor value calculated for the desired output frequency. (Default = 13h)
287
®
R
Address offset:
06004h–06007h
Default:
00100053h
Attributes:
R/W
Size:
32 bits
The Display Clock 0 Divisor Register and Display & LCD Clock Divisor Select Register are programmed with the loop
31
26
Reserved
Bit
25
VCO 1 N-divisor
16
15
10
Reserved
9
0
VCO 1 M-divisor
Description
31-26
Reserved
25-16
15-10
Reserved
9-0
288
®
R
Address offset:
06008h–0600Bh
Default:
00030013h
Attributes:
R/W
Size:
32 bits
The Display Clock 2 Divisor Register and Display & LCD Clock Divisor Select Register are programmed with the loop
31
26
Reserved
Bit
25
VCO 2 N-divisor
16
15
10
Reserved
9
0
VCO 2 M-divisor
Description
31-26
Reserved
25-16
15-10
Reserved
9-0
289
®
R
14.5 LCD_CLKD — LCD Clock Divisor Register
Address offset:
0600Ch–0600Fh
Default:
00030013h
Attributes:
R/W
Size:
32 bits
The LCD Clock Divisor Register and Display & LCD Clock Divisor Select Register are programmed with the loop
31
26
Reserved
Bit
25
16
VCO LCD N-divisor
15
10
Reserved
9
VCO LCD M-divisor
Description
31:6
Reserved
25:16
VCO LCD N-divisor. N-divisor value calculated for the desired output frequency. (Default = 03h)
15:10
Reserved
9:0
290
0
VCO LCD M-divisor. M-divisor value calculated for the desired output frequency. (Default = 13h)
®
R
14.6 DCLK_0DS — Display & LCD Clock Divisor Select Register
Address offset:
06010h–06013h
Default:
40404040h
Attributes:
R/W
Size:
32 bits
Display clock i {i=0 to 2} becomes effective after the appropriate byte i {i = 0 to 2} in this register is programmed. The
LCD clock becomes effective after byte 3 in this register is programmed.
31
30
Reserved
23
Post divisor select LCD clk
22
Reserved
15
20
Post divisor select clk 2
14
Reserved
7
28
12
Post divisor Select clk 1
6
Reserved
4
Post divisor select clk 0
Bit
31
30:28
27
26
Reserved
VCO
loop div
LCD clk
19
18
Reserved
VCO
loop div
clk 2
11
10
Reserved
VCO
loop div
clk 1
3
2
Reserved
VCO
loop div
clk 0
25
24
Reserved
17
16
Reserved
9
8
Reserved
1
0
Reserved
Description
Reserved
Post Divisor Select LCD Clock
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
291
®
Bit
R
Description
27
Reserved
26
VCO Loop Divide LCD Clock
0 = Divided by 4*M (default) (M = LCD Clock Divisor Register [9:0])
1 = Divided by 16*M
25:23
Reserved
22:20
Post Divisor Select clock 2
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
19
Reserved
18
VCO Loop Divide clock 2
1 = Divided by 16*M
17
Reserved
16
Reserved
15
Reserved
14:12
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
11
Reserved
10
1 = Divided by 16*M
9:7
Reserved
6:4
000 = Divide by 1
001 = Divide by 2
010 = Divide by 4
011 = Divide by 8
101 = Divide by 32
11x = Reserved
3
292
Reserved
®
R
Bit
2
Description
1 = Divided by 16*M
1:0
Reserved
293
®
R
14.7 PWR_CLKC — Power Management and Miscellaneous Clock
Control
Address offset:
6014h–06017h
Default:
00 00 01 03 h
Attributes:
R/W
Size:
32 bits
31
17
16
Reserved
Reserved
15
12
Reserved
11
10
9
Unit
level clk
gating for
3D
cluster
disable
Unit
level clk
gating for
2D
cluster
disable
Unit
level
clock
gating for
MI
cluster
disable
7
2
Reserved
Bit
31:17
16
15:12
11
Description
Reserved
Reserved (0)
Reserved
Unit-level clock gating for 3D cluster disable
0 = Unit-level clock gating for all units in 3D enabled (default)
1 = Unit-level clock gating for all units in 3D disabled
10
Unit-level clock gating for 2D cluster disable
0 = Unit-level clock gating for all units in 2D enabled (default)
1 = Unit-level clock gating for all units in 2D disabled
294
8
Reserved
1
0
Display
clock
PLL
VCO
Internal
DAC
enable
®
R
Bit
9
Description
Unit-level clock gating for MI cluster disable
0 = Unit-level clock gating for all units in MI enabled (default)
1 = Unit-level clock gating for all units in MI disabled
Notes:
8
7:2
1
Reserved 1
Reserved
Display Clock PLL VCO
0 = Disable
1 = Enable (default)
0
Internal DAC Enable
0 = Disables the internal DAC (PowerDown). If HSYNC/VSYNCControl[0] = 0, disables HSYNC and VSYNC.
1 = Enables the internal DAC and does not allow disable of HSYNC and VSYNC via HSYNC/VSYNC Control[0]
(default).
295
®
R
296
®
R
15. Video Registers
This section contains the Video Overlay and Gamma Correction registers and an Overlay instruction. The current
graphics controller implements one overlay, which is referred to as Overlay 0. Note that the Overlay 0 control registers
are indirectly written by first setting up a buffer in memory and then instructing the graphics controller to update the onchip registers from this buffer. Software can invoke the update process by writing to the OV0ADD register or by issuing
the Overlay Flip instruction. Note that the Gamma Correction registers are read/written directly. The register/instruction
categories are listed in the Overlay Register/Instruction Categories table.
System Memory
Graphics Controller
Register Range
(512 KB)
Base+301xxh
Overlay 0 Reg
On-chip registers for Overlay 0
(read only; debug)
(Base = MMADR PCI Reg.)
Graphics Memory
Base+xxh
(Base = OV0ADD Reg.)
Overlay 0 Reg
Memory buffer area for loading on-chip registers
- Software setsup register values
- HW updates on-chip regs for next VBLANK
overlay1.vsd
Do not use the “Wait for VBLANK” mechanism to force a sequence of overlay flips. Use the “Wait for Scan Lines”
mechanism with the scan line set up to be at least 1 scan line after vertical blank start, in order to force the loading of the
next Overlay x Register Update Address, which will take effect after the next displayed overlay frame.
297
®
R
Table 11. Overlay Register/Instruction Categories
Register/Instruction Category
Overlay 0 Register Update Address
(OV0ADD)
Comment
Mem. Address
Offset
30000h
Used to update Overlay 0 registers.
Provides physical memory address of buffer area
used for updating on-chip registers.
Write of OV0ADD register causes hardware to
update on-chip registers on next VBLANK.
Display/Overlay 0 Status Register
(DOV0STA)
30008h
Overlay Status bits
Gamma Correction (GAMMA[0:5])
30010h-30027h
Not part of double-buffer scheme. Access
registers directly.
Overlay Register Sets
301xxh
(on chip RO
regs)
“xx” indicates a particular register address.
Base+xxh
(Mem. buffer
regs)
Software sets up buffer in memory.
Overlay Buffer Pointer
Overlay Stride
Overlay Initial Phase
Overlay Dest. Window Position/Size
Overlay Source Size
Overlay Scale Factor
Overlay Color Correction
Overlay Destination Color Key
On-chip registers are not directly writeable.
Registers are double buffered (buffer in
memory and on-chip registers).
Software writes to OV0ADD Register, which
provides memory buffer address location
and causes hardware to read memory buffer
and update on-chip registers during next
VBLANK.
On-chip registers can be read through debug read
path. See OV0ADD register description.
Overlay Source Color Key
Overlay Configuration
Overlay Command
Overlay Flip Instruction
298
NA
This instruction invokes the Overlay register
update.
®
R
15.1 OV0ADD—Overlay 0 Register Update Address Register
Memory address offset:
30000h–30003h
Default value:
00000000
Access:
R/W
Size:
32 bits
This register provides a physical memory address that will be used on the next register update for Overlay 0. This
register is double buffered to allow it to be updated during overlay display.
Updating Register Values
A write to this register sets an internal bit (readable through the status register) that causes all register values written to
the memory buffer area to be loaded into the corresponding on-chip registers and to become active on the next VBLANK
event. Overlay 0 Flip is asserted after all registers are updated from memory.
31
30
Memory
Select
29
Reserved
Bit
31
28
0
Register Update Address
Description
Memory Select. The registers must be in the same memory as the overlay data that it describes. Intel
82810 Chipset = Overlay is always in non-cacheable system memory (1), never in local memory.
0 = Local memory
1 = Non-cacheable system memory
30-29
28:0
Reserved.
Register Update Address. Physical memory address that will be used on the next register update for
Overlay 0.
299
®
R
15.2 DOV0STA—Display/Overlay 0 Status Register
30008h–3000Bh
Default value:
TBD
Access:
RO
Size:
32 bits
This read-only register indicates the status for the overlay. References to display are either the primary timing generator
or the secondary timing generator, depending on which is currently being used.
31
30
24
Reg Update
Status
Reserved
23
21
19
15
14
13
12
Reserved
Not Active
Pixel
Reserved
Not Active
Video Scan
Line
Bit
31
20
Overlay 0 Current
Buffer/Field.
Reserved
11
Reserved
18
17
16
Reserved
Reserved
Reserved
10
0
Display Line Status.
Description
Register Update Status. All registers latched (flip pending = 0).
0 = Overlay update register has been written, and VBLANK event and registers have not been loaded
from memory.
1 = Update registers have not changed since the last VBLANK active edge.
30:21
Reserved
20:19
Overlay 0 Current Buffer/Field. This field indicates the current buffer. Updated at display VBLANK
before interrupt. Field is valid only when in field mode.
00 = Buffer 0 Field 0
18
Reserved
17
Reserved
16
Reserved
15
Reserved
300
®
R
Bit
14
Description
Not Active Pixel. This bit indicates the Display Horizontal Blank Active state (including border). Updated
in real time, it is set by the leading edge of the overlay’s display HBLANK and is cleared by the trailing
edge of the overlay’s HBLANK.
0 = HBlank inactive
1 = HBlank active
13
Reserved
12
Not Active Video Scan Line. This bit indicates the Display Vertical Blank Active state (including border).
Updated in real time, it is set by the leading edge of the overlay’s display VBLANK and is cleared by the
trailing edge of the overlay’s VBLANK.
0 = VBlank inactive
1 = VBlank active
11
10:0
Reserved
Display Line Status. This field displays the line number. Reset to zero at the trailing edge of display
VBLANK. Incremented at the trailing edge of the overlay’s display HBLANK.
301
®
15.3 Gamma Correction
Note that the registers in this section are read/written directly at the memory address offset location specified.
15.3.1
GAMC[5:0]—Gamma Correction Registers
30010h–30027h
GAMC5 = 30010h
GAMC4 = 30014h
GAMC3 = 30018h
GAMC2 = 3001Ch
GAMC1 = 30020h
GAMC0 = 30024h
Default value:
GAMC5 = C0C0C0h
GAMC4 = 808080h
GAMC3 = 404040h
GAMC2 = 202020h
GAMC1 = 101010h
GAMC0 = 080808h
Access:
R/W
Size:
32 bits
These registers determine the characteristics of the gamma correction for the overlay data. Each register is 32 bits, which
are written and read together when accessed from the PCI.
The GAMCxx registers are not double buffered and should only be updated when the overlay engine is disabled. During
operation, the bytes of these registers are read independently, depending on the R, G or B pixel values.
302
R
®
R
31
24
Reserved
23
16
15
Red
Bit
8
Green
7
0
Blue
Description
31:24
Reserved.
23:16
Red.
15:8
Green.
7:0
Blue.
GAMC5: Used of bit <7:6> = 11
GAMC4: Used of bits <7:6> = 10
303
®
15.3.2
R
Mathematical Gamma Correction for Overlay
Gamma correction is a function that corrects for nonlinearity of the display phosphor brightness as a function of the
electron beam current. Depending on the actual phosphor used and/or the case that data streams have been pre-gamma
corrected for different phosphor characteristics than the display device being used, the digital display stream must be
corrected to achieve equally spaced increase in brightness on the display for equally spaced increases in color intensity
values.
Gamma correction is implemented via a piecewise linear approximation of a curve. There are 6 individual breakpoint
values, logarithmically spaced in the color intensity domain. Only one set of 8 values is provided and shared for each of
the red, green, and blue intensity components. These values are programmable via software control. For each color red,
green, and blue, the appropriate gamma breakpoint pairs are looked up and smoothly interpolated, in order to arrive at
the final red, green, and blue values that are output to the DAC.
The inputs to the function are:
• PCI register bus: The chip-internal PCI data bus and the appropriate register decodes for loading the gamma
breakpoint values
• Red In 2e07:00: The red color component of the overlay stream
• Green In 2e07:00: The red color component of the overlay stream
• Blue In 2e07:00: The red color component of the overlay stream
The output of the function is:
• Red out:
The gamma-corrected red color component
• Green out:
The gamma-corrected green color component
• Blue out:
The gamma-corrected blue color component
15.3.2.1
Gamma Correction Theory of Operation
The gamma correction function outlined above is implemented via piecewise linear interpolation that allows any
arbitrary curve to be output as a function of any linear input. The function provides smooth (continuous) interpolations
efficiently in hardware, while compromising little in the precision of the output curve relative to the ideal. The
breakpoints of the curve are defined via software and loaded via the PCI internal data bus.
In general, the implementation outlined above allows for 7 logarithmically spaced breakpoints, of which the 5 internal
breakpoints are software definable, and the 2 endpoints are predefined at 0 and 255 (decimal). The value at a given
breakpoint and the breakpoint+1 (using the MSB, which is 1 of an 8-bit color) are looked up in the table simultaneously.
These two points define the exact, desired values at those breakpoints. The values between the breakpoints are linearly
interpolated using the remaining bits of the color. The output of the function therefore evaluates the equation:
Result = mX + B , where:
m = slope
B = initial value defined at table entry N
The slope, m, is derived by taking the difference between the breakpoint defined by [N] and the next successive
breakpoint [N+1]. The value at [N+1] is simply derived via the input wiring to the mux. Since all slopes are defined to be
positive for this function, the use of an unsigned multiplier is sufficient for this function.
304
®
Latch Output Value
R
Latch Address
Area Of Interest
The following is a more detailed description of the algorithm. A line with two coordinate points represented by (x1, y1)
and (x2, y2) can be defined by the following equation:
y2 = slope (x2 - x1) + y1
This implies that the slope of the line is:
slope = (y2 - y1) / x2 - x1
Another point with X = x3 on the same line can be defined as (x3, y3), where y3 is
y3 = slope (x3 - x1) + y1
Derived Slope =
Adrs[N+1] - Adrs[N]
Interpolated Point
Value Of Latch
@ Adrs [N+1]
Ideal Curve
Value Of Latch
@ Adrs [N]
Value Of LSBs
Result = Value Of Latch @ Adrs[N] + Slope * LSBs
Gamma correction can be bypassed by programming the registers with data values corresponding to a linear curve with
slope = 1. The register programming for gamma bypassing is as shown below.
Adrs[8] = 8
Adrs[64] = 64
Adrs[16] = 16
Adrs[128] = 128
Adrs[32] = 32
Adrs[192] = 192
When registers are programmed with the above values, the output of the gamma unit is the same as the input, so no
gamma correction is performed.
305
®
15.3.2.2
R
Gamma Hardware Implementation
Result = Adrs[192] + [[0x00
- Adrs[192]] * [X - 192] << 1] >> 7 192 <= X < 256
Result = Adrs[128] + [[Adrs[192] - Adrs[128]] * [X - 128] << 1] >> 7 128 <= X < 192
Result = Adrs[64 ] + [[Adrs[128] - Adrs[ 64 ]] * [[X - 64] << 1]] >> 7 64 <= X < 128
Result = Adrs[32 ] + [[Adrs[64 ] - Adrs[ 32 ]] * [[X - 32] << 2]] >> 7 32 <= X < 64
Result = Adrs[16 ] + [[Adrs[32 ] - Adrs[ 16 ]] * [[X - 16] << 3]] >> 7 16 <= X < 32
Result = Adrs[ 8 ] + [[Adrs[16 ] - Adrs[ 8 ]] * [[X - 8] << 4]] >> 7 8
<= X < 16
Result = 0x00
<= X < 8
+ [[Adrs[8
]-
0x00 ]] * [[X - 0 ] << 4]] >> 7 0
Consider Adrs[8] = 8, Adrs[16] = 16, Adrs[32] = 32, Adrs[64] = 64, Adrs[128] = 128, Adrs[192] = 192.
For X = 5, result = 0 + [[8-0] * [5 << 4]] >> 7
== 5
For X = 11, result = 8 + [[16-8] * [3 << 4]] >> 7
== 11
For X = 24, result = 16 + [[32-16] * [8 << 3]] >> 7
== 24
For X = 63, result = 31 + [[64-32] * [31 << 2]] >> 7
== 63
For X = 100, result = 64 + [[128-64] * [36 << 1]] >> 7
== 100
For X = 156, result = 128 + [[192-128] * [28<< 1]] >> 7
== 156
For X = 200, result = 192 + [[256-192] * [8 << 1]] >> 7
== 200
15.4 Overlay Buffer Pointer Registers
These registers provide address pointers into the system memory or local-memory buffer areas. The buffers must be
qword aligned. Pixel panning on a pixel basis is done using the byte addresses. Overlay buffers need to be qword
aligned, and the stride should be a qword multiple. Buffer pointers should always be aligned to the natural boundaries
based on the data format. For planar formats (YUV410, YUV420), the Y and UV pointers should be naturally aligned to
each other. Their natural alignment depends on the particular data format.
Pixel
Format
Alignment
Pixels
Bytes
RGB packed
1
2
YUV 4:2:2 packed
2
4
YUV 4:1:1 packed
8
12
YUV Planar
1
1
Only the Register Update Address register should be written while the overlay is active. Otherwise, values will be loaded
by writing the register image into memory and writing the command register with the address of the memory image in
the Register Update address.
306
®
R
15.4.1
OBUF_0Y—Overlay Buffer 0 Y Pointer Register
Memory buffer address offset:
00h (R/W)
On-chip reg. mem. addr. offset:
30100h (RO; debug path)
Default value:
00h
Access:
See address offset above.
Size:
32 bits
31
26
25
0
Reserved
Overlay buffer 0 Y pointer
Bit
Description
31:26
Reserved
25:0
For Y planar or packed color data (byte address). Must be pixel aligned (low-order bit zero for 16-bpp
packed formats). When mirroring horizontally (X backwards), this points to the last byte of the line.
15.4.2
OBUF_1Y—Overlay Buffer 1 Y Pointer Register
04h (R/W)
Default value:
00h
Access:
Size:
32 bits
31
26
25
Reserved
0
Overlay buffer 1 Y pointer
Bit
Description
31:26
Reserved
25:0
Overlay Buffer 1 Y Pointer. For Y Planar or packed color data (byte address). Must be pixel aligned (loworder bit zero for 16-bpp packed formats). When mirroring horizontally (X backwards), this points to the
last byte of the line.
15.4.3
OBUF_0U—Overlay Buffer 0 U Pointer Register
08h (R/W)
307
®
R
Default value:
00h
Access:
Size:
32 bits
31
26
25
0
Reserved
Overlay buffer 0 U pointer
Bit
Description
31:26
Reserved
25:0
Overlay Buffer 0 U Pointer. YUV planar modes only. UV address in the interleaved UV formats (byte
address)
15.4.4
OBUF_0V—Overlay Buffer 0 V Pointer Register
0Ch (R/W)
3010Ch (RO; debug path)
Default value:
00h
Access:
Size:
32 bits
31
26
Reserved
Bit
25
0
Description
31:26
Reserved
25:0
Overlay Buffer 0 V Pointer. YUV non-interleaved planar formats only (byte address)
308
®
R
15.4.5
OBUF_1U—Overlay Buffer 1 U Pointer Register
10h (R/W)
Default value:
00h
Access:
Size:
32 bits
31
26
25
0
Reserved
Bit
Description
31:26
Reserved
25:0
Overlay Buffer 1 U Pointer. YUV Planar Modes only. UV address in the interleaved UV formats (byte
address)
15.4.6
OBUF_1V—Overlay Buffer 1 V Pointer Register
14h (R/W)
Default value:
00h
Access:
Size:
32 bits
31
26
Reserved
Bit
25
0
Description
31:26
Reserved
25:0
Overlay Buffer 1 V Pointer. YUV non-interleaved planar formats only (byte address)
309
®
R
15.5 Overlay Stride Registers
These values represent the width of the buffer that contains the overlay data. This is independent of the actual width that
gets displayed and is used to determine the line-to-line increment of the buffer. In two-line buffer mode, there is only
room for 180 quadwords per scan line. If the source address is not quadword aligned, then for formats YUV4:2:0,
YUV4:2:2, and RGB, the source must be less then 720 pixels. The stride must be quadword aligned.
15.5.1
OV0STRIDE—Overlay 0 Stride Register
Default value:
Access:
Size:
31
29
Reserved
Bit
18h (R/W)
00h
32 bits
28
16
Overlay 0 UV planar buffer
stride
15
13
Reserved
12
0
Overlay 0 Y planar or
YUV/RGB buffer stride
Description
31:29
Reserved
28:16
Overlay 0 UV planar Buffer Stride. Only used for YUV planar formats and gives the U or V buffer stride
in bytes. This is a two’s complement number and is negative during Y mirroring. Low-order three bits are
always zero, forcing a qword alignment. The range is ±4 KB.
15:13
Reserved
12:0
Overlay 0 Y planar or YUV/RGB Buffer Stride. Buffer (Y planar or YUV/RGB packed) stride in bytes.
This is a two’s complement number and is negative during Y mirroring. Low-order three bits are always
zero. The range is ±4 KB.
310
®
R
15.6 Overlay Initial Phase Registers
Provides a spatial sub-pixel-accurate adjustment. This value is always a fractional positive number such that, when
combined with the subtract one from initial phase bit, the possible range for initial phase becomes –1 < phase < 1. There
two separate vertical initial-phase registers are used in field-based image processing.
15.6.1
YRGB_VPH—Y/RGB Vertical-Phase Register
Default value:
Access:
Size:
31
20
Y/RGB Vertical-Phase
Buffer/Field 1
1Ch (R/W)
00h
32 bits
19
16
Reserved
Bit
31:20
15
Y/RGB Vertical-Phase
Buffer/Field 0
4
3
0
Reserved
Description
Y/RGB Vertical-Phase Buffer/Field 1. This fractional value sets the initial vertical phase.
In packed formats:
YUV/RGB data buffer 1 when up-scaling in frame mode
YUV/RGB data field 1 when up-scaling in field mode
In planar YUV formats:
Y data buffer 1 when up-scaling in frame mode
Y data field 1 when up-scaling in field mode
19:16
Reserved
15:4
Y/RGB Vertical-Phase Buffer/Field 0. This fractional value sets the initial vertical phase.
In packed formats:
YUV/RGB data buffer 0 when up-scaling in frame mode
YUV/RGB data field 0 when up-scaling in field mode.
In planar YUV modes:
Y data buffer 0 when up-scaling in frame mode
Y data field 0 when up-scaling in field mode.
3:0
Reserved
311
®
15.6.2
R
UV_VPH—UV Vertical-Phase Register
Default value:
Access:
Size:
31
20
20h (R/W)
00h
32 bits
19
UV Vertical Phase 1
16
15
Reserved
4
3
UV Vertical Phase 0
0
Reserved
Bit
Description
31:20
UV Vertical Phase 1. This fractional value is used only in YUV planar formats where the UV plane may
have a different vertical initial phase from the Y data. This field is used with buffer 1 in frame mode or Field
1 in field mode.
19:16
Reserved
15:4
UV Vertical Phase 0. This fractional value is used only in YUV planar formats where the UV plane may
have a different vertical initial phase from the Y data. This field is used with buffer 0 in frame mode or field
0 in field mode.
3:0
Reserved
15.6.3
HORZ_PH—Horizontal-Phase Register
Default value:
Access:
Size:
31
20
UV Horizontal Phase
Bit
24h (R/W)
00h
32 bits
19
16
Reserved
15
4
Y/RGB Horizontal Phase
3
0
Reserved
Description
31:20
UV Horizontal Phase. Sets the initial horizontal phase for the UV data. Only used in YUV modes.
19:16
Reserved
15:4
Y/RGB Horizontal Phase. Sets the initial horizontal phase for both buffers and fields. Unlike the vertical
initial phases, this does not change buffer to buffer or field to field. YUV modes use a separate initial
phase for Y and UV data. This value will either be the actual initial phase or the initial phase minus one.
3:0
Reserved
312
®
R
15.6.4
INIT_PH—Initial Phase Register
Default value:
Access:
Size:
28h (R/W)
00h
32 bits
31
6
Reserved
Bit
5
0
Initial Phase minus one
Description
31:6
Reserved
5:0
5
Initial Phase Minus One. These bits provide a method of creating a negative initial phase. If the
corresponding bit is set, the initial phase is the register value minus one. These bits should only be set in
cases where the buffer pointer is pointing to the first pixel of the line or column, because it will effectively
cause the first pixel to be duplicated.
4
Y Vertical Buffer / Field 0
3
Y Vertical Buffer / Field 1
2
Y Horizontal
1
UV Vertical Buffer / Field 0
0
UV Vertical Buffer / Field 1
UV Horizontal
313
®
R
15.7 Overlay Destination Window Position/Size Registers
These registers allow for the positioning of the overlay data relative to the graphics display or the secondary display’s
active region. It allows pixel-accurate positioning. When a secondary display is used, the area outside the overlay
windows will be black RGB(0,0,0).
15.7.1
DWINPOS—Destination Window Position Register
Default value:
Access:
Size:
31
27
Reserved
Bit
2Ch (R/W)
00h
32 bits
26
16
Display vertical top line
15
11
Reserved
10
0
Display horizontal left pixel
Description
31:27
Reserved
26:16
Display Vertical top line. Display vertical top in lines 0 = begin at display first line.
15:11
Reserved
10:0
Display Horizontal left pixel. Determines where in the display screen coordinates the overlay display will
begin.
Display horizontal left in pixels 0 = begin at display left edge.
314
®
R
15.7.2
DWINSZ—Destination Window Size Register
Default value:
Access:
Size:
31
27
Reserved
Bit
30h (R/W)
00h
32 bits
26
16
Vertical Size
15
11
Reserved
10
0
HorizontalSize
Description
31:27
Reserved
26:16
Vertical Size. Destination vertical size in lines (never specifies scan lines off the active display)
15:11
Reserved
10:0
Horizontal Size. Destination horizontal size in pixels (never specifies pixels off the active display)
315
®
R
15.8 Overlay Source Size Registers
These registers provide information to the overlay engine regarding what data needs to be fetched from memory. If the
overlay destination window is smaller than the result of the scaled-up source, it will be clipped on the right and bottom of
the overlay window. The source data is clipped within 1 qword. The source width must not specify an additional qword
or more of data that is not required to satisfy the destination. If the scaled source is smaller than the destination window,
the last pixel or last scan line accessed is used to fill the right and bottom areas. 2 Qwords is the minimum source width
per operand (Y, U, and V for planar formats; packed formats have only 1 operand).
15.8.1
SWID—Source Width Register
Default value:
Access:
Size:
31
24
Reserved
34h (R/W)
00h
32 bits
23
16
UV Source Width
Bit
15
9
Reserved
8
0
Y/RGB Source Width
Description
31:24
Reserved
23:16
UV Source Width. The number of bytes contained in a single line of planar UV source data. This is
unused in packed modes. For planar modes, it’s used for the U and V source width (which are assumed to
be the same) or in the interleaved mode (YI64) as TBD. When the last pixel is reached and the complete
destination window has not been filled, this pixel will be repeated until the end of the destination window.
When displaying YUV 4:1:1 data, this field contains the number of U bytes, which is identical to the
number of V bytes and ¼ of the Y bytes.
15:9
Reserved
8:0
Y/RGB Source Width. The number of bytes contained in a single line of source data. In planar modes,
this is the Y source width.
This should include all contributing pixel data.
When the last pixel is reached and the complete destination window has not been filled, this pixel will be
repeated until the end of the destination window.
When displaying YUV 4:2:2 data, the atomic unit is a doubleword (2 Ys + U + V).
When displaying YUV 4:1:1 data, the atomic unit is 3 doublewords (8 Ys + 2U + 2V).
The starting offset within the buffer must reflect this restriction.
316
®
R
15.8.2
SWIDQW—Source Width in Qwords Register
Default value:
Access:
Size:
31
24
Reserved
Bit
38h (R/W)
00h
32 bits
23
16
UV Source Width in QWs
15
9
Reserved
8
0
Y/RGB Source Width in
QWs
Description
31:24
Reserved
23:16
UV Source Width in Qwords. The number of QWs contained in a single line of planar UV source data.
Minimum size is 2 qwords.
15:9
Reserved
8:0
Y/RGB Source Width in Qwords. The number of QWs contained in a single line. In planar modes, this is
the Y source width. Minimum size is 2 qwords.
317
®
15.8.3
R
SHEIGHT—Source Height Register
Default value:
Access:
Size:
31
26
Reserved
Bit
3Ch (R/W)
00h
32 bits
25
16
UV Source Height
15
11
Reserved
10
0
Y/RGB Source Height
Description
31:26
Reserved
25:16
UV Source Height. In packed formats this is unused. For planar YUV formats, it indicates the number of
lines, starting at and including the base address line contained in the UV source data. When the last line is
reached and the complete destination window has not been filled, this line will be repeated until the end of
the destination window.
A height must not be specified for lines that are not on the active display.
Minimum height is 2 lines or 4 lines for interlaced surfaces (Bob).
15:12
Reserved
10:0
Y/RGB Source Height. In packed formats, this indicates the number of lines, starting at and including the
base address line contained in the source data. In planar formats, it is the number of Y lines. This is used
to determine where the end of the source is in the vertical direction, in order to handle the edge effects
related to the vertical filter. When the last line is reached and the complete destination window has not
been filled, this line will be repeated until the end of the destination window.
A height must not be specified for lines that are not on the active display.
Note: In the BOB (interlaced) case for video-capture flip-mode (when the overlay acts as a slave to videocapture), the height must be an EVEN number, in order to ensure that field 0 and field 1 have an equal
number of source lines.
The minimum height is 2 lines or 4 lines for interlaced surfaces (Bob).
318
®
R
15.9 Overlay Scale Factor Registers
These registers provide the scaling information used to specify the amount of vertical and horizontal scaling. In the case
of YUV formats, there are independent scale factors for Y and UV data, in order to compensate for the various formats
that include sub-sampled UV data. Up- or down-scaling is set using the bits of the command word. In the case of downscaling, the integer portion of the scale factor can be up to three in the vertical and is assumed to be one in the horizontal.
15.9.1
YRGBSCALE—Y/RGB Scale Factor Register
Default value:
Access:
Size:
31
20
Vertical Scale Factor
Y/RGB
Bit
40h (R/W)
00h
32 bits
19
17
Reserved
16
15
Horizontal
Scale Factor
Y/RGB Integer
14
X Scale Factor Y/RGB
3
2
Reserved
1
0
Vertical
Down-Scale
Integer
Description
31:20
Vertical Scale Factor Y/RGB. This fractional positive number represents the scale factor to be used in
vertical scaling. For packed formats, it applies to all color components. For planar YUV formats, it is used
for the Y component only. Always 1 > scale > 0 (1/scale for up-scale fractional portion of the scale factor
for down-scaling).
19:17
Reserved
16:15
Horizontal Scale Factor Y/RGB Integer. This field is used only for horizontal down-scale. In the case of
planar formats, it specifies the integer portion of the scale factor for Y data. Only horizontal down-scaling
of 2 (or less) to 1 is supported.
14:3
Horizontal Scale Factor Y/RGB. This fractional positive number represents the scale factor to be used in
horizontal scaling. Always 1 > scale ≥ 0.
2
1:0
Reserved
Vertical Down-Scale Integer. This field is used only for vertical down-scale. In the case of planar formats,
it specifies the integer portion of the scale factor for Y data.
319
®
15.9.2
R
UVSCALE—U V Scale Factor Register
Default value:
Access:
Size:
44h (R/W)
00h
32 bits
31
20
19
Vertical Scale Factor UV
15
Reserved
16
Reserved
14
3
Horizontal Scale Factor UV
2
Reserved
1
0
UV Down-Scale
Integer
Bit
Description
31:20
Vertical Scale Factor UV. Used only in YUV planar modes to set the scale factor for the UV data. This is
different in formats where the UV data is vertically sub-sampled.
19:15
Reserved
14:3
Horizontal Scale Factor UV. Used only in YUV modes to set the scale factor for the UV data. This is
different in formats where the UV data is sub-sampled.
2
1:0
320
Reserved
UV Down-Scale Integer. This field is used only for vertical down-scale. It specifies the integer portion of
the scale factor for UV data.
®
R
15.10 Overlay Color Correction Registers
Used for YUV sources only. Adjustments are made before the RGB conversion.
15.10.1 OV0CLRC0—Overlay 0 Color Correction 0 Register
Default value:
Access:
Size:
48h (R/W)
00h
32 bits
31
17
16
Reserved
8
7
Contrast
Bit
0
Brightness
Description
31:17
Reserved
16:8
Contrast. - 7.3F (3.6 format)
Contrast is bypassed, even for RGB, by programming this field to 1.0.
Brightness. Range: ±127. A value of zero disables this adjustment affect. This value gets added to the Y
value after contrast multiply and just before RGB conversion.
7:0
Brightness is bypassed, even for RGB, by programming this field to 0.
15.10.2 OV0CLRC1—Overlay 0 Color Correction 1 Register
Default value:
Access:
Size:
4Ch (R/W)
00h
32 bits
31
10
Reserved
Bit
31:10
9:0
9
0
Saturation/CrCb
Description
Reserved
Saturation/CrCb. This operates in two modes: When YUV data is being operated on, the register
contains the saturation value. When CbCr data is being used, it is the sum of the saturation multiplier
value added to the CbCr scale factor (128/112).
This unsigned, fixed-point number is in 3.7 format.
Saturation is bypassed, even for RGB, by programming this field to 1.0
321
®
R
15.11 Overlay Destination Color Key Registers
Used for YUV sources only. Adjustments are made before RGB conversion.
15.11.1 DCLRKV—Destination Color Key Value Register
Default value:
Access:
Size:
31
24
50h (R/W)
00h
32 bits
23
0
Reserved
Destination Color Key Value
Bit
Description
31:24
Reserved
23:0
Destination Color Key Value. In the format of the destination (screen). Usable only when the screen is
used as the destination (not TV). When the overlay is directed to TV output, this value replaces pixels that
have been suppressed by the source color key.
15.11.2 DCLRKM—Destination Color Key Mask Register
Default value:
Access:
Size:
54h (R/W)
00h
32 bits
31
30
29
Dest
Color Key
En
Dest Const α
Blend Enable
Always Const
α Blend
Enable
322
28
24
Reserved
23
0
Destination Color Key mask
®
R
Bit
31
Description
Destination Color Key Enable
1 = Destination color key is enabled.
0 = Destination color key is disabled.
30
Destination Constant Alpha Blend Enable
1 = Enable constant alpha blending between the overlay and the primary display when the destination
color key does not match, and destination color keying is enabled within the alpha blend window.
0 = Disable constant alpha blending.
29
Always Constant Alpha Blend Enable
1 = Enable constant alpha blending within the alpha blend window.
0 = Disable constant alpha blending.
28:24
Reserved
23:0
Destination Color Key Mask
0 = Bits that are active participants in the compare
1 = Bits that are active participants in the compare. A mask of all ones will disable the color key (as if all
colors match).
15.12 Overlay Source Color Key Registers
There is an overlay source key per overlay stream that is used on a pixel basis. Source comparison occurs after horizontal
zooming, but in the YUV formats before color space conversion. If the source data (overlay) is within the range, then the
primary display is selected.
If the overlay is in RGB mode, the most-significant bits are duplicated on the least-significant, in order to form 8-bit
channels:
R<4:0> → R<4:0 ‘ 4:2>
G<5:0> → G<5:0 ‘ 5:4>
B<4:0> → B<4:0 ‘ 4:2>
Before the source key comparison is made…
Destination color key failing takes precedence over the source chroma key failing. If both fail, then the primary display
is selected.
323
®
R
15.12.1 SCLRKVH—Source Color Key Value High Register
Default value:
Access:
Size:
31
24
58h (R/W)
00h
32 bits
23
0
Reserved
Source Key value High
Bit
Description
31:24
Reserved
23:0
Source Key Value High. This value is the high value that is compared with the overlay pixel per 8-bit
channel. An overlay value greater than this field on any enabled channel fails the comparison and passes
the overlay pixel.
15.12.2 SCLRKVL—Source Color Key Value Low Register
Default value:
Access:
Size:
31
24
5Ch (R/W)
00h
32 bits
23
Reserved
Bit
0
Source Key value Low
Description
31:24
Reserved
23:0
Source Key Value Low. In the format of the source, specifies the low end (greater than or equal to) of the
range of excluded source pixel data. The software sets bits that it does not want to be included in the
comparison to 0.
1 = Included in comparison
0 = Not included in comparison
324
®
R
15.12.3 SCLRKM—Source Color Key Mask Register
Default value:
Access:
Size:
31
30
Source
Const α
Blend
Enable
27
Reserved
60h (R/W)
00h
32 bits
26
24
23
Source Key Mask
Enables
Bit
31
0
Constant α
Red<7:0>
Constant α
Green <7:0>
Constant α
Blue<7:0>
Description
Source Constant Alpha Blend Enable
1 = Enable source alpha blending when the logical OR of the source key mask enables are asserted
within the alpha blend window, and the comparison indicates that the overlay is to be displayed.
0 = Disable source alpha blending.
30:27
Reserved
26:24
Source Key Mask Enables. Each bit enables one channel. If the bit is a one, then the comparison result
is used. Otherwise, it is ignored.
Bit 2 = Enables comparison [23:16]
23:16
Constant Alpha Red <7:0>
pixel (R) = (alphaR * primary display (R)) + ((1-alphaR) * overlay (R))
Intel 82810 Chipset implementation:
This involves three 4x8-bit alpha multipliers to be inserted in the overlay and primary display merging
logic. When the alpha blend is selected, the most-significant 4 bits of each alpha channel are used as the
alpha term. If the original 8 bits of the alpha channel = FF, then the alpha channel value is treated as a 1.
If alpha = 0, then 1-0 must also be treated as a 1. This functionality only works for a primary display of 16
and 24 bits per pixel.
15:8
Constant Alpha Green <7:0>
pixel (G) = (alphaG * primary display (G)) + ((1-alphaG) * overlay (G))
325
®
R
Bit
Description
Constant Alpha Blue <7:0>
7:0
pixel (B) = (alphaB * primary display (B)) + ((1-alphaB) * overlay (B))
15.13 Overlay Configuration Registers
There is only 1 Overlay Configuration register, which controls both overlay streams. It is read from memory with
Overlay 0 register loads during vertical blank.
15.13.1 OV0CONF—Overlay Configuration Register
Default value:
Access:
Size:
64h (R/W)
00h
32 bits
31
8
Reserved
7
6
5
4
Reserved
Cr Cb
Rdjust En
(Reserved)
Reserved
YUV Conv
En
(Reserved)
Bit
Description
31:7
Reserved
6
Reserved
5
Reserved
4
Reserved
3:1
Reserved
0
Line Buffer Configuration. Sets the line buffer configuration:
0 = two 720 pixel line buffers
1 = one 1440 pixel line buffer
326
3
1
Reserved
0
Line Buf
Conf
®
R
15.14 OV0CMD—Overlay Command Register
Default value:
Access:
Size:
68h (R/W)
00h
32 bits
This register provides the data that the overlay engine needs to begin its work. A write to this register sets an internal bit
(readable by the status) that will cause all the register values that were written to be internally latched and to become
active on the next VBLANK event.
31
30
28
Sel. Top OV
(Reserved)
27
25
Vertical Chrom Filter
23
22
21
19
Horiz. Chrom Filter (cont.)
15
14
18
Horizontal Luminance Filter
17
Mirroring
13
10
4:2:2: Byte Order
24
Horiz Chrom
Filter [24:22]
Vertical Luminance Filter
Source Format
7
6
5
4
3
Flip Qual
(Cont)
Vert. Initial
Phase Sel.
Disp Flip
Type
Ignore Buf
and Field
Reserved
16
Y Adj.
9
8
Flip TVOut
Field Sel.
Flip Qual
[8:7]
2
1
Buffer and Field
0
Overlay
Enable
327
®
Bit
31
30:28
27:25
24:22
21:19
18:17
16
328
R
Description
Select top overlay. Reserved for future implementations
Vertical Chrominance Filter. Vertical chrominance filter
000 = Scaling off (1:1)
001 = Line replication
010 = Up interpolation
011 = Reserved
100 = Reserved
101 = Pixel dropping
110 = Down interpolation
111 = Reserved
Vertical Luminance Filter. Vertical luminance filter
011 = Reserved
100 = Reserved
111 = Reserved
Horizontal Chrominance Filter. Horizontal chrominance filter
011 = Reserved
100 = Reserved
111 = Reserved
Horizontal Luminance Filter. Horizontal luminance filter also applies to RGB.
011 = Reserved
100 = Reserved
111 = Reserved
Mirroring. Mirroring affects the buffer address values used. See the buffer address description for details.
00 = Normal
01 = Horizontal mirroring
10 = Vertical mirroring
11 = Both horizontal and vertical
Y adjust. Defines the range of Y and UV values in the source data. This bit is only has an effect in YUV
formats.
0 = Y 0-255 UV ± 128
1 = Y 16-235 UV ± 112
UV source values are always in excess 128 format.
®
R
Bit
15:14
13:10
9
8:7
6
Description
4:2:2 Byte Order. Affects the byte order for 4:2:2 data. For other data formats, it should be set to zero.
00 = Normal
01 = UV swap
10 = Y swap
11 = Y and UV swap
Source Format.
0000 = Reserved
0001 = Reserved
0010 = RGB 5:5:5
0011 = RGB 5:6:5
0100 = Reserved
0101 = Reserved
0110 = Reserved
0111 = Reserved
1000 = YUV 4:2:2
1001 = YUV 4:1:1
1010 = Reserved
1011 = Reserved
1100 = YUV 4:2:0 (MPEG-1 or 2)
1101 = Reserved
1110 = YUV 4:1:0
1111 = Reserved
Flip TV Out Field Select. Selects the TV-out field polarity for flips.
0 = Between F0 and F1
1 = Between F1 and F0
Flip Qualification. Flip qualification (& display VBLANK). Flips can be caused automatically by the capture
port logic on the completion of a field or frame capture. This event will be synchronized either with the
display/overlay VBLANK event or the display/overlay VBLANK event combined with the current TV field.
Manual flips can occur as the result of an update of the overlay registers due to the writing of the Overlay
Update Address Register when the register data specifies that the buffer/field should be changed or
through a command packet that specifies a register update that changes the buffer/field. This also can be
synchronized with the TV field.
Manual flip command
00 = Flip (standard)
01 = Flip & TV-out Field #
Automatic flipping
10 = Capture frame/field
11 = Capture frame/field & TV-out field #
11 = Reserved encoding for the Intel 82810 Chipset
Vertical Initial Phase Select. Selects the initial vertical-phase register to use. The choice is to always use
the same register or to alternate (for field processing) between the two registers.
This bit will be overridden by the capture port when autoflipping.
0 = Use only field/buffer 0 initial vertical phase
1 = Use both initial vertical-phase values
329
®
R
Bit
Description
5
Display/Flip Type. This bit affects the buffer addressing used for buffer display and the use of the initial
vertical phase. Frame mode starts addressing at the value contained in the buffer address register and
increments by the stride as it increments from line to line. Initial phase selection is based on the buffer and
the vertical initial-phase select bit.
Field mode uses the field bit to determine if the start address should be the value in the start address
register or the start address register plus stride. Field mode will increment the address by twice the stride
as it increments from line to line. Initial phase selection is based on the field and the vertical initial phase
select.
This bit will be overridden by the capture port when autoflipping.
0 = Frame Mode
1 = Field Mode
Ignore Buffer and Field. When this field is set, don’t update the buffer and field through the command
register.
0 = Use buffer and field data to update buffer/field
1 = Don’t update buffer/field
Buffer and field. Selects the displayed buffer and field. This determines which buffer and field will be
displayed when the overlay is enabled or when the ignore bit was clear. It would otherwise be ignored and
the internal buffer/field values would be used. These are readable through the status register.
Overlay Enable. Overlay Enable
0 Disable (no display or memory fetches)
1 Enable
Changing this bit from a 0 to 1 will cause the overlay to begin display after the next qualified flip event. A
disable (1 → 0) will cause the overlay to stop displaying and image on this current display VBLANK.
4
2:1
0
330
®
R
15.15 Overlay Alpha Blend Window Position/Size Registers
These registers allow for the alpha blending of a subsection of the overlay window positioning of the overlay data
relative to the graphics display or the secondary display active region. It allows pixel-accurate positioning. The Overlay
Alpha Blend Window must be programmed to be either equal to or within the Overlay Window.
15.15.1 AWINPOS—Alpha Blend Window Position Register
Default value:
Access:
Size:
31
27
Reserved
70h (R/W)
00h
32 bits
26
16
Alpha Blend Vertical top
line
Bit
15
11
Reserved
10
0
Alpha Blend Horizontal left
pixel
Description
31:27
Reserved
26:16
Alpha Blend Vertical top line.
Determines where, on the display screen coordinates, the overlay display is alpha blended.
Alpha blend vertical top, in lines (0 = begin at display first line)
15:11
Reserved
10:0
Alpha Blend Horizontal left pixel.
Determines where, on the display screen coordinates, the overlay display are alpha blended.
Alpha blend horizontal left, in pixels (0 = begin at display left edge)
331
®
R
15.15.2 AWINSZ—Alpha Blend Window Size Register
Default value:
Access:
Size:
31
27
74h (R/W)
00h
32 bits
26
Reserved
16
Alpha Blend Vertical Size
Bit
15
11
Reserved
10
0
Alpha Blend Horizontal
Size
Description
31:27
Reserved
26:16
Alpha Blend Vertical Size.
Alpha blend vertical size, in lines (never specifies scan lines off the active display)
15:11
Reserved
10:0
Alpha Blend Horizontal Size.
Alpha blend horizontal size, in pixels (never specifies pixels off the active display)
15.16 Overlay Flip Instruction
DWord
0
1
Bit
Description
31:29
Client: xxh
28:23
Function Index:
22:16
Instruction Target
31:0
Register Update Address:
Do not use the “Wait for VBLANK” mechanism to force a sequence of overlay flips. Use the “Wait for Scan Lines”
mechanism, with the scan line set up to be at least 1 scan line after vertical blank start, in order to force the loading of the
next overlay x register update address, which will take effect after the next displayed overlay frame. For the Intel 82810
Chipset, the Overlay Update Address register can be loaded either before or after the “Wait for VBLANK” (primary
display VBLANK).
332
®
R
16. Instruction and Interrupt Control Registers
16.1 Instruction Control Registers
16.1.1
FENCE—Graphics Memory Fence Table Registers
Address offset:
Default value:
Access:
Size:
02000h - 0201Fh
00000000h
Read/32-bit, Write Only
8×32 bits each
The Memory Hub (MH) performs address translation between linear space and tiled space. The intent of tiling is to
locate graphics data that are close (in X and Y surface axes) in one memory page, while still locating some amount of
line-oriented data sequentially in memory for display efficiency. All 3D rendering is done in such a manner that all
qwords of any one span are located in the same memory page, which improves rendering performance.
Tiled memory is supported for rendering surfaces located in graphics memory. A tiled memory surface is a surface with
a secondary pitch and height that are subsets of the surface’s total pitch and height. The graphics controller maintains the
constants required by the memory interface in order to perform the address translations for up to eight tiled regions (each
with a different pitch and size).
In order to perform address translation, the memory interface needs the surface pitch and tile height. It uses the fence
base address, size constants, and surface address in order to determine if the rendering surface is tiled and to swizzle the
bits as required by tiling. Since the fence ranges are at least 512 KB and aligned, 7 address bits are required to specify
the lower bound. In addition, the fence lower bound must be fence size aligned.
A tile represents 2 KB of memory. The tile height is fixed at 16. Based on this, the surface pitch must be programmed in
tiles. Eight fence table registers occupy the address range specified above. Each fence table register has the following
format.
Engine restrictions on tile surface usage are detailed in the Cspec. Note that X and Y major tiles are used for host,
texture, and blitter source and destination surfaces. The display, overlay, capture, motion comp source, dest surfaces,
color and Z surfaces (if tiled) must be X major.
31
26
25
19
Reserved
15
Reserved
Fence Lower Bound
14
13
Reserved
12
11
Tile Walk
Reserved
18
16
Reserved
10
8
Fence Size
333
®
7
R
6
Reserved
4
3
Fence Pitch
Bit
1
Reserved
Description
31:26
Reserved for address bits 31 to 26.
25:19
Fence Lower Bound : Memory address bits 25 to 19 (must be size aligned)
18:15
Reserved
14:13
Reserved: MBZ (00)
12
11
10:8
7
Reserved
Fence size:
000 = 512 KB
001 = 1 MB
010 = 2 MB
011 = 4 MB
100 = 8 MB
101 = 16 MB
110 = 32 MB
111 = Reserved
Reserved
6:4
Fence Pitch:
000 = 1 tile
001 = 2 tiles
010 = 4 tiles
011 = 8 tiles
100 = 16 tiles
101 = 32 tiles
110 = Reserved
111 = Reserved
3:1
Reserved
0
334
Tile walk:
0 = X major
1= Y major
Fence Valid:
0 = Unused
1 = Valid
0
Fence
Valid
®
R
16.1.2
PGTBL_CTL—Page Table Control Register
Address offset:
Default value:
Access:
Size:
02020h
00000000h
Read/Write
32 bits
This register enables/disables the page table mechanism. When enabled, it also sets the base address of the 4-KB-aligned
page table.
The driver writes this register when it creates the page table at the start of virtual mode. A 64-KB physical contiguous
region in memory-mapped space is remapped to the page table located in local memory. Driver writes to this memorymapped space are offset into the page table, as defined by the page table base. The physical address will be the
combination of the page table base and the page table offset derived from the memory mapped driver write. A translation
resource access (TLB access) with the page table disabled results in an interrupt.
The page table can be disabled when there is no drawing engine (render-map/blitter) or high priority stream
(overlay/display/capture) active. The hardware incorporates several TLBs that cache page table entries. Disabling the
page table through this register will invalidate all TLBs, except the ones serving display and overlay. The following table
lists all TLBs and their invalidation mechanisms:
TLB
Normal Invalidation Mechanism
Note
Display
Refreshed on Vsync
Not affected by bit 0
Overlay
Refreshed on Vsync
Mot affected by bit 0
Render/Blit
Flush
Invalidated by page table disable
Host
Through a page table write
Mapping
Command Stream
335
®
31
R
12
11
Page Table Base Address
(4-KB aligned)
Bit
1
Reserved
Page Table Base Address is 4-KB aligned. Must be within main memory.
11:1
Reserved
Page Table Enable: If the graphics memory range is accessed through the graphics translation table
when this bit is not set, an interrupt event is generated. All graphics streams other than cursor and VGA
fall within this category.
0 = Disable
1 = Enable
336
Pg Tbl
Enable
Description
31:12
0
0
®
R
16.1.3
PGTBL_ER—Page Table Error Register (Debug)
Address offset:
Default value:
Access:
Size:
02024h
0000000000h
Read Only
32 bits
This register stores information pertaining to page table error interrupts. An invalid physical address implies a region in
main memory with restricted accessibility (e.g., PAM/SMM space).
The display engine speculatively fetches data and may cross GTT-mapped pages that are not intended for use. To
prevent errors and the hanging of the display engine, GTT pages adjacent to the contiguous display region must be left
unused but marked valid and mapped.
31
12
Physical Address 31:12
11
6
Reserved
5
Error ID
3
2
0
Error Type
Bit
Description
31:12
Physical Address. This field provides the page table access address in case of an invalid address. Not
valid for other error types.
11:6
Reserved
5:3
Error Identification: Identifies the TLB that caused the error. After an error, render, mapping and blitter
engines will stop executing. Also, capture will stop and may result in overrun. However, overlay, display,
and host operations will not stop. Each source records the first error and ignores subsequent errors.
000 = Video capture
001 = Overlay
010 = Display
011 = Host
100 = Render
101 = Blitter
110 = Mapping
111 = Command stream
Error Type:
000 = Invalid table
001 = Invalid page table entry
010 = Incorrect target for display surface (Request to lm if the surface started in mm or vice-versa) /
overlay surface (request to lm).
011 = Invalid miss during display/overlay accesses
100 = Illegal translation data (Translation is valid and address points to pam, smm, over top, and other
restricted spaces in main memory.)
101 = Access to local memory when not present
110 = Surface tiled in Y when not allowed (render/display/overlay).
111 = Reserved
2:0
337
®
16.1.4
R
RINGBUF—Ring Buffer Registers
Address offset:
02030h - 0207Fh
02030h - 0203Fh: Low-priority ring
02040h - 0204Fh: Interrupt ring
02050h - 0205Fh: Reserved
02060h - 0207Fh: Reserved
00000000h
Read/32-bit, Write Only
4 dwords
Default value:
Access:
Size:
Each ring buffer is defined by a 4-dword register set that includes the starting address, length, head pointer, and tail
pointer. The ring buffer can be disabled when the ring is empty. Drivers uses two sets. The format of the ring buffer
states is as follows.
Dword
Offset
Bit
0
31:21
20:3
2:0
31:2
1
2
3
1:0
31:26
25:12
11:0
31:21
20:12
11:3
2:1
Description
Reserved
Tail Pointer : Programmable qword offset in the ring buffer (20:3 is used by the hardware)
Reserved
Head pointer: Hardware-maintained dword offset in the ring buffer (20:2 is used by the
hardware. Bits 31:21 are incremented whenever the head pointer wraps from the end to the
start of the ring buffer <> wrap count.)
Reserved
Reserved
Starting Address: Programmable 4-KB page aligned address of the buffer
Reserved
Reserved
Buffer Length: Programmable length of the ring buffer in 4-KB Pages (max. = 2 MB, x000
= one 4-KB page)
Reserved
Automatic Report Head Pointer: Report happens when the ring DMA crosses 64-KB or
128-KB boundary.
X0 = No report
01 = Report every 16 pages (64 KB)
11 = Report every 32 pages (128 KB)
0
Ring Buffer Valid:
0 = Disabled
1 = Enabled
338
®
R
16.1.5
HWS_PGA—Hardware Status Page Address Register
Address offset:
Default value:
Access:
Size:
02080h
1FFFF000h.
Read/Write
32 bits
Hardware status page physical address. The programmed address should be 4-KB aligned. Bits [11:0] are hardwired to 0.
This page is used to report hardware status into system memory, as follows.
Note that for the GMCH, bits 31:29 must be 0.
31
29
Must be 0s
28
11
Base physical address of status page
[28:12]
Dword
Offset
0
0s (hardwired)
Description
0
Report Interrupt Status Register
1
Report Low Priority Ring Head Pointer
2
Report Interrupt Ring Head Pointer
3
Reserved for other ring
(1K-1): 4
12
Can be written through indexed store dword instruction.
339
®
16.1.6
R
IPEIR—Instruction Parser Error Identification Register (Debug)
Address offset:
Default value:
Access:
Size:
02088h
0000h
Read Only
32 bits
This register is used to help identify the instruction packet that generates an invalid instruction interrupt to the CPU. The
IPEIR contains the origin of the offending packet. The header is stored in the error header register.
31
3
Reserved
Bit
31:3
2
1:0
16.1.7
2
1
Batch /
ring
0
Ring ID
Description
Reserved
Batch / Ring. The invalid instruction came from either a batch buffer or instruction ring.
1 = Batch buffer
0 = Instruction ring
Ring ID. The invalid instruction came from the low-priority (00) or the interrupt (01) ring. If the
invalid instruction came from a batch buffer, this bit identifies the instruction ring from which the
batch buffer instruction came.
00 = Low-priority
01 = Interrupt ring
1X = Reserved
IPEHR—Instruction Parser Error Header Register (Debug)
Address offset:
Default value:
Access:
Size:
0208Ch
0000h
Read Only
32 bits
This register is used to identify the instruction packet that generates an invalid instruction interrupt to the CPU. The
IPEHR contains the header word of the offending packet. For debug purposes, the headers of all parsed instructions will
be written to this register. If an error occurs, the instruction parser halts. An interrupt indicating an error will be
generated if it is unmasked.
Bit
Description
31:0
Header. This field contains the instruction header field of the instruction packet that generates
an invalid instruction interrupt.
340
®
R
16.1.8
INSTDONE—Instruction Stream Interface Done Register
Address offset:
Default value:
Access:
Size:
02090h
FFFFh
Read Only
16 bits
This read only register reports engine done signals.
15
8
Reserved for unit done signals (TBD)
7
6
5
4
3
2
1
0
Reserved
Blitter
done
Mapping
eng. done
Render
eng. done
Batch
done
Reserved
Intr. ring
empty
Low-prior.
ring emp.
Bit
15:8
Description
Reserved for unit done signals (TBD)
7
Reserved
6
Blitter Done
5
Mapping Engine Done
4
Render Engine Done
3
Batch Done
2
Reserved
1
Intr. Ring Empty or Disabled
0
Low-Priority Ring Empty or Disabled
341
®
16.1.9
R
NOPID—NOP Identification Register
Address offset:
Default value:
Access:
Size:
02094h
00000000h
Read Only
32 bits
This register contains the value specified by the last GFXCMDPARSER_NOP_IDENTIFICATION instruction received.
31
24
23
0
Reserved
Identification Number
Bit
Description
31:22
Reserved
21:0
Identification Number
16.1.10 INSTPM—Instruction Parser Mode Register
Address offset:
Default value:
Access:
Size:
020C0h
00h
Read/Write
8 bits
The bits in this register control the operation of the instruction parser.
NOTE: If an instruction type is disabled, the parser will read it out of the instruction / batch FIFO, but will not send it to
its destination. Error checking will be performed.
342
7
6
5
4
3
2
1
0
Reserved
Enable
Sync
Packet
AGP Flush
(Reserved)
Enable
Sync
Packet
Flush
Disable
Mcomp
Instrs
Disable
GDI Blitter
Instrs
Disable
Render
(3D/Stretch)
Instrs
Disable
State
Variable
Updates
Disable
Render
Palette
Updates
®
R
Bit
Description
7
Reserved
6
Enable Sync AGP Flush. Enable pipe and AGP flush. Set by software and cleared by the parser on
detecting graphics pipe flushed before parsing a subsequent packet.
1 = Enable
0 = Cleared by graphics controller
5
Enable Sync flush. Enable pipe flush. Set by software and cleared by the parser on detecting graphics
pipe flushed before parsing a subsequent packet.
1 = Enable
0 = Cleared by graphics controller
4
Disable Mcomp Instructions. Disable processing of mcomp instructions by parser.
1 = Disable
0 = Enable
3
Disable GDI Blitter Instructions. Disable processing of blitter instructions.
1 = Disable
0 = Enable.
2
Disable Render (3D/Stretch) Instructions. Disable processing of 3D instructions (client 00h) by parser.
1 = Disable
0 = Enable
1
Disable State Variable Updates
1 = Disable
0 = Enable
0
Disable Render Palette Updates
1 = Disable
0 = Enable
343
®
R
16.1.11 INSTPS—Instruction Parser State Register (Debug)
Address offset:
Default value:
Access:
Size:
020C4h
0000h
Read Only
32 bits
This register contains the state code of the instruction parser in the CSI. Decoding the contents of this register will
indicate what the instruction parser is doing currently.
Bit
31:0
Description
Instruction Parser State
16.1.12 BBP_PTR—Batch Buffer Parser Pointer Register (Debug)
Address offset:
Default value:
Access:
Size:
020C8h
00000000h
Read Only
32 bits
This register contains the offset from the batch buffer start address of the dword being parsed by the instruction parser.
31
19
Reserved
Bit
2
Address Offset
Description
31:19
Reserved
18:2
Batch Buffer Address Pointer Offset
1:0
Reserved
344
18
1
0
Reserved
®
R
16.1.13 ABB_STR—Active Batch Buffer Start Address Register (Debug)
Address offset:
Default value:
Access:
Size:
020CCh
00000000h
Read Only
32 bits
This register is loaded with the start address of the Batch Buffer request.
The ABB_STR and ABB_END registers will not get loaded if they are popped off of the stack (i.e., if a low-priority ring
batch buffer is interrupted at a chain point by interrupt priority ring execution and then continued later). The start and
end addresses for the low-priority ring chain portion that was interrupted will not be stored in the ABB_STR and
ABB_END registers. This operational anomaly will not be corrected.
31
26
25
Reserved
Batch buffer request start address
Bit
Reserved
25:3
Batch Buffer Request Start Address
1:0
2
Reserved
1
0
Source of batch
buffer
Description
31:26
2
3
Reserved
Source of the Batch Buffer
00 = Low-priority ring
01 = Interrupt ring
1X = Reserved
345
®
R
16.1.14 ABB_END—Active Batch Buffer End Address Register (Debug)
Address offset:
Default value:
Access:
Size:
020D0h
00000000h
Read Only
32 bits
This register is loaded with the end address of the batch buffer request.
The ABB_STR and ABB_END registers will not get loaded if they are popped off of the stack (i.e., if a low-priority ring
batch buffer is interrupted at a chain point by interrupt priority ring execution and then continued later). The start and
end addresses for the low-priority ring chain portion that was interrupted will not be stored in the ABB_STR and
ABB_END registers. This operational anomaly will not be corrected.
31
26
Reserved
25
Batch buffer request end address
Bit
Description
31:26
Reserved
25:3
Batch Buffer Request End Address
2
1:0
346
3
Reserved
Source of Batch Buffer
01 = Interrupt ring
1X = Reserved
2
Reserved
1
0
Source of batch
buffer
®
R
16.1.15 DMA_FADD—DMA Engine Fetch Address (Debug)
Address offset:
Default value:
Access:
Size:
020D4h
00000000h
Read Only
32 bits
This register contains the offset from the start address of the instruction being fetched by the DMA engine.
31
26
25
Reserved
Bit
Reserved
25:3
Current DMA Address
1:0
Current address of DMA pointer
2
Reserved
1
0
User of DMA
engine
Description
31:26
2
3
Reserved
User of the DMA Engine:
01 = Interrupt ring
10 = Reserved
11 = Batch
347
®
R
16.2 Interrupt Control Registers
All interrupt control registers described below share the same bit definition, as follows:
15
14
13
Reserved
HW Detect
Error
Master
12
11
10
9
8
Sync Status
Toggle
Pri Dply Flip
Pending
Sec Dply
Flip
Pending
(Rsvd in
GMCH)
Overlay 0
Flip
Pending
Overlay 1
Flip
Pending
(Rsvd in
GMCH)
7
6
5
4
3
2
1
0
Pri Dply
VBLANK
Pri Dply
Event
Sec Dply
VBLANK
(Rsvd in
GMCH)
Sec Dply
Event
(Rsvd in
GMCH)
Host Port
Event
(Rsvd in
GMCH)
Capture
Event
(Rsvd in
GMCH)
User
Defined
Interrupt
Breakpoint
Table 12. Bit Definition for Interrupt Control Registers
Bit
Description
15
Hardware Detected Error Master. When this status bit is set, it indicates that the hardware has detected
an error. It is set on an error condition and cleared by a CPU write of a 1 to the appropriate bit contained
in the error ID register, followed by a write of a 1 to this bit in the IIR. Further information on the source of
the error comes from the Error Status Register that, along with the Error Mask Register, determines which
error conditions will cause the error status bit and the interrupt to occur. The intent is to use the error bits
to detect errors during debug and testing. Error conditions during normal operation should not occur.
MM/LM refresh timer error: Indicates a refresh request buffer overrun.
Page Table Error: Indicates a page table error.
Display or overlay underrun: Set on either a display or overlay underrun error. See display and overlay
status registers to determine source of the error.
Capture FIFO overrun: Set on a capture buffer overrun. (Reserved. Not implemented in the GMCH.)
Host port error: Set on a host port error. (Reserved. Not implemented in the GMCH.)
Instruction parser error : The instruction parser encounters an error while parsing an instruction.
14:13
Reserved
12
Sync Status Toggle. This bit is toggled when the instruction parser completes a flush with the sync
enable bit active in the instruction parser mode register. The toggle event will happen after all graphics
engines are flushed. The store dword resulting from this toggle also will cause the CPU’s view of graphics
memory to be coherent (i.e., invalidates the host graphics cache).
11
Primary Display Flip Pending. Status bit is set on a pending flip and cleared when the flip occurs,
whereas the IIR reflects the Flip-Occurred#, which is contrary to the general definition of the setting of IIR
bits when interrupts occur. This is used only when the GFXCMDPARSER_FRONT_BUFFER _INFO
packet is being used. See that instruction for additional information. To prevent race conditions, the status
write occurs before the STOREDWORD following the flip packet is written.
10
Secondary Display Flip Pending (Reserved. Not implemented in the GMCH.)
348
®
R
Bit
Description
9
Overlay 0 Flip Pending. The status bit is set to reflect a pending flip when the parser parses a flip packet,
and it is cleared when the flip takes place (display VBLANK), whereas the IIR reflects the Flip-Occurred#,
which is contrary to the general definition of setting of IIR bits when interrupts occur. This is only affected
by the use of flip packets, not through the manual method or capture auto flipping. To prevent race
conditions, the status register write must occur before the STOREDWORD following the flip packet is
written.
8
Overlay 1 Flip Pending (Reserved. Not implemented in the GMCH.)
7
Primary Display VBLANK. Set at the leading edge of display VBLANK. This is actually delayed to allow
all internal hardware VBLANK events to occur before the interrupt is generated, in order to eliminate race
conditions. These events include the update of the display and overlay status bits and the loading of the
overlay registers.
6
Primary Display Event. The interrupt cause will be determined by reading the display status register, and
the cause is one of the following:
Flat panel hot plug detect interrupt
Display VSYNC
Display line compare
On active-going edge of OR of unmasked display event bits
Status: OR of unmasked display event bits
Note that the display line compare also is used through the instruction parser packet interface.
5
Secondary Display VBLANK (Reserved. Not implemented in the GMCH.) Set at the leading edge of
display VBLANK. This is actually delayed in order to allow all VBLANK events to occur before the interrupt
is generated. These events include the update of the overlay registers.
4
Secondary Display Event (Reserved. Not implemented in the GMCH.)
3
Host Port Event (Reserved. Not implemented in the GMCH.)
2
Capture Event (Reserved. Not implemented in the GMCH.)
Interrupt: OR of the unmasked capture interrupt events
Status: OR of the unmasked capture events
1
User-Defined Interrupt. The instruction parser passed a “user-defined interrupt” packet. This is intended
for use with another mechanism (e.g., STOREDW instruction), in order to determine the source of the
packet.
0
Breakpoint. The instruction parser parsed a “breakpoint” interrupt packet. The
GFXCMDPARSER_BREAKPOINT_INTERRUPT packet can be used to generate a store dword cycle
from the command streamer to main memory and to halt the parsing of further commands. The HWSTAM
(Hardware Status Mask Register, offset: 02098h) must have the breakpoint bit unmasked in order to
generate the store dword. In addition to the HWSTAM register, the corresponding bit in the IMR (Interrupt
Mask Register, offset: 020A8h) must also be unmasked in order to halt the parsing of further commands.
At the breakpoint command packet, if both the HWSTAM and IMR are unmasked, the command parser
will stop processing further commands until the breakpoint bit in the IIR (Interrupt Identity Register, offset:
020A4h) is cleared.
349
®
16.2.1
R
HWSTAM—Hardware Status Mask Register
Address offset:
Default value:
Access:
Size:
02098h
FFFFh
Read/Write
16 bits
This register has the same format as the Interrupt Control Registers. The corresponding bits are the mask bits that prevent
that bit in the Interrupt Status Register from generating a PCI write cycle. Any unmasked interrupt bit (set to 0) will
allow the Interrupt Status Register to be written to the address specified by the Hardware Status Vector Address Register,
when the Interrupt Status Register changes state.
15
14
13
Reserved
HW detect
error
master
12
11
10
9
8
Sync status
toggle
Pri dply flip
pending
Sec dply flip
pending
(rsvd in
GMCH)
Overlay 0
flip pending
Overlay 1
flip pending
(rsvd in
GMCH)
7
6
5
4
3
2
1
0
Pri dply
VBLANK
Pri dply
event
Sec. dply
VBLANK
(rsvd in
GMCH)
Sec. dply
event: (rsvd
in GMCH)
Host port
event: (rsvd
in GMCH)
Capture
event (rsvd
in GMCH)
Userdefined
interrupt
Breakpoint
Bit
15:0
Description
Interrupt Status Mask Bits.
0 = Not masked
1 = Masked (prevents PCI write cycle)
350
®
R
16.2.2
IER—Interrupt Enable Register
Address offset:
Default value:
Access:
Size:
020A0h
0000h
Read/Write
16 bits
Individual enables for each interrupt described above. A disabled interrupt will still appear in the Interrupt Identity
Register, to allow the polling of interrupt sources.
15
14
13
HW Detect
Error
Master
Reserved
11
10
9
8
Sync Status
Toggle
Pri Dply Flip
Pending
Sec Dply
Flip
Pending
(Rsvd in
GMCH)
Overlay 0
Flip
Pending
Overlay 1
Flip
Pending
(Rsvd in
GMCH)
7
6
5
4
3
2
1
0
Pri Dply
VBLANK.
Pri Dply
Event
Sec Dply
VBLANK
(Rsvd in
GMCH)
Sec Dply
Event:
(Rsvd in
GMCH)
Host Port
Event:
(Rsvd in
GMCH)
Capture
Event
(Rsvd in
GMCH)
User
Defined
Interrupt
Breakpoint
Bit
15:0
12
Description
Interrupt Enables. (See Table 12.)
1 = Enable.
0 = Disable.
351
®
16.2.3
R
IIR—Interrupt Identity Register
Address offset:
Default value:
Access:
Size:
020A4h
0000h
Read/Write Clear
16 bits
The individual interrupt(s) that occurred are determined via this register. The bit is set by the interrupt event and held
until cleared by writing a 1 into the bit position.
15
14
13
HW Detect
Error
Master
Reserved
10
9
8
Sync Status
Toggle
Pri Dply Flip
Pending
Sec Dply
Flip
Pending
(Rsvd in
GMCH)
Overlay 0
Flip
Pending
Overlay 1
Flip
Pending
(Rsvd in
GMCH)
6
5
4
3
2
1
0
Pri Dply
VBLANK.
Pri Dply
Event
Sec Dply
VBLANK
(Rsvd in
GMCH)
Sec Dply
Event:
(Rsvd in
GMCH)
Host Port
Event:
(Rsvd in
GMCH)
Capture
Event
(Rsvd in
GMCH)
User
Defined
Interrupt
Breakpoint
Description
Interrupt Identity. See Table 12.
1 = Interrupt occurred.
352
11
7
Bit
15:0
12
®
R
16.2.4
IMR—Interrupt Mask Register
Address offset:
Default value:
Access:
Size:
020A8h
FFFFh
Read/Write
16 bits
An interrupt masked by this register will not appear in the Interrupt Identity Register and will not generate an interrupt.
15
14
13
HW Detect
Error
Master
Reserved
11
10
9
8
Sync Status
Toggle
Pri Dply Flip
Pending
Sec Dply
Flip
Pending
(Rsvd in
GMCH)
Overlay 0
Flip
Pending
Overlay 1
Flip
Pending
(Rsvd in
GMCH)
7
6
5
4
3
2
1
0
Pri Dply
VBLANK.
Pri Dply
Event
Sec Dply
VBLANK
(Rsvd in
GMCH)
Sec Dply
Event:
(Rsvd in
GMCH)
Host Port
Event:
(Rsvd in
GMCH)
Capture
Event
(Rsvd in
GMCH)
User
Defined
Interrupt
Breakpoint
Bit
15:0
12
Description
Interrupt Mask Bits. See. Table 12
0 = Not masked
1 = Masked
353
®
16.2.5
R
ISR—Interrupt Status Register
Address offset:
Default value:
Access:
Size:
020ACh
0100h (probably still not exactly the correct value)
Read Only
16 bits
This register contains the non-persistent value of the signal causing each interrupt. These bits are not masked by the
Interrupt Mask Register. The user interrupt and the breakpoint interrupt generate one-clock pulses. The corresponding
bits in this register will serve no practical purpose due to the short duration of the signal.
15
14
13
HW Detect
Error
Master
Reserved
10
9
8
Sec Dply
Flip
Pending
(Rsvd in
GMCH)
Overlay 0
Flip
Pending
Overlay 1
Flip
Pending
(Rsvd in
GMCH)
6
5
4
3
2
1
0
Pri Dply
VBLANK.
Pri Dply
Event
Sec Dply
VBLANK
(Rsvd in
GMCH)
Sec Dply
Event:
(Rsvd in
GMCH)
Host Port
Event:
(Rsvd in
GMCH)
Capture
Event
(Rsvd in
GMCH)
User
Defined
Interrupt
Breakpoint
Description
Interrupt Status. See Table 12.
1 = Signal caused an interrupt.
354
11
Pri Dply Flip
Pending
7
Bit
15:0
12
Sync Status
Toggle
®
R
16.2.6
Error Identity, Mask and Status Registers
The Error Identity, Mask, and Status registers have the following bit descriptions. The master error bit in the ISR and IIR
register will be set when the OR of the unmasked (with a zero in the corresponding mask register bits) error status bits is
TRUE.
16.2.6.1
®
Page Table Error Handling in Intel 810 Chipset
In the following cases, page table errors can be caused by accessing graphics aperture space:
1) When page table is not enabled
2) During access to local memory in a UMA system
3) During access to a page table entry that does not have the valid bit set
4) During the location of the page table in a restricted region (e.g., SMM space) of memory
For cycles initiated from the graphics host, TLB error should not cause the system to hang.
TLB error is flagged only for a write. For the write cycle with the TLB error and for all subsequent write cycles, byte
enables are masked. If local memory is accessed in the UMA system, the cycle is forwarded to system memory and
completed with masked byte enables. For a write TLB error, the error registers should be set appropriately.
Graphics memory reads do not cause any side-effects. Hence, all reads with TLB error are allowed to complete. Note
that no error condition will be set in the error registers for a read TLB error.
Regardless of whether the cycle is a read or a write, if the page table is located in a restricted region, a page table error
will be set. This error is not resettable.
16.2.6.2
Resetting the Page Table Error
The page table error will be reset every time a write cycle is generated to bit 15 of the Interrupt Identity register (IIR),
regardless of the setting of the bit in the IMR or the IER. Resetting the page table error should cause the subsequent write
cycles to be completed without masking of the byte enables. Please note that it is impossible to clear a TLB error
resulting from an illegal location of the page in system memory.
355
®
15
6
Reserved
R
5
4
3
2
1
0
MM/LM
refresh
timer error
Page table
error
Display or
overlay
underrun
Capture
overrun
(rsvd in
GMCH)
Host port
error (rsvd
in GMCH)
Instruction
parser error
Bit
15:6
5
Description
Reserved
MM/LM Refresh Timer Error:
1 = Refresh overrun
4
Page Table Error:
1 = Page table error
0 = Cleared by a CPU write of a 1 to the error identity bit.
3
Display or Overlay Underrun:
1 = Display or overlay underrun error
0 = Cleared by a CPU write of a 1 to the Error Identity bit. To determine the source of the error, see the
display and overlay status registers.
2
®
Capture Overrun (reserved in the Intel 810 chipset). Set on a capture buffer overrun.
1 = Overrun
1
®
Host Port Error (reserved in the Intel 810 chipset). Set on a host port error.
1 =Error
0
Instruction Parser Error: The instruction parser encounters an error while parsing an instruction.
1 = Error
0 = Cleared by a CPU write of a 1 to the Error Identity bit.
356
®
R
16.2.6.3
EIR—Error Identity Register
Address offset:
Default value:
Access:
Size:
020B0h
00h
Read/Write Clear
16 bits
15
6
Reserved
5
4
3
2
1
0
MM/LM
refresh
timer error
Page table
error
Display or
overlay
underrun
Capture
overrun
(rsvd in
GMCH)
Host port
error (rsvd
in GMCH)
Instruction
parser error
Bit
Description
Error Identity Bits. (See the error identity, mask, and status register bit definition table.)
15:0
1 = Error occurred.
16.2.6.4
EMR—Error Mask Register
Address offset:
Default value:
Access:
Size:
020B4h
FFh
Read/Write
16 bits
15
6
Reserved
Bit
15:0
5
4
3
2
1
0
MM/LM
refresh
timer error
Page table
error
Display or
overlay
underrun
Capture
overrun
(rsvd in
GMCH)
Host port
error (rsvd
in GMCH)
Instruction
parser error
Description
Error Mask Bits. (See the error identity, mask, and status register bit definition table.)
0 = Not masked
1 = Masked
357
®
16.2.6.5
ESR—Error Status Register
Address offset:
Default value:
Access:
Size:
020B8h
FFh
Read Only
16 bits
15
6
Reserved
Bit
15:0
358
R
5
4
3
2
1
0
MM/LM
refresh
timer error
Page table
error
Display or
overlay
underrun
Capture
overrun
(rsvd in
GMCH)
Host port
error (rsvd
in GMCH)
Instruction
parser error
Description
Error Status Bits. (See the error identity, mask, and status register bit definition table.)
®
R
17. LCD / TV-Out Register Description
During the LCD or TV-Out mode, the BIOS will program the following LCD / TV-Out registers, which are 32-bit,
memory-mapped registers. These registers are not double-buffered and take effect when loaded. Further, this subsystem
takes into account modified CR register values during the vertical blank time for centering.
This subsystem allows the timing generator to be programmed to pixel granularity. The only exception is during the
VGA pixel doubling mode. During VGA pixel doubling, the active pixel time must be a multiple of four pixels to
account for centering with VGA pixel doubling, and non-active times must be a multiple of 2 pixels clocks.
All fields are excess-0 encoded. This means that the hardware uses the value+1, where the value is the entry in the field.
Therefore, if a 0 is programmed into a field, the hardware uses the value 1 for that field.
17.1 HTOTAL — Horizontal Total Register
Address offset:
60000h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
Reserved
27
16
Horizontal total display pixels
Bit
15
11
Reserved
10
0
Horizontal active display
pixels
Description
31:28
Reserved. Read Only
27:16
Horizontal Total Display Pixels
This 12-bit field provides a horizontal total up to 4096 pixels, which encompassing 2048 active display
pixels, front/back border pixels, and the horizontal retrace period. Any pending event (HSYNC, VSYNC) is
reset at htotal.
15:11
Reserved. Read Only
10:0
Horizontal Active Display Pixels
This 11-bit field provides horizontal active display resolutions up to 2048 pixels. Note that the first
horizontal active display pixel always starts at 0.
359
®
R
17.2 HBLANK — Horizontal Blank Register
Address offset:
60004h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
27
Reserved
16
Horizontal blank end
Bit
15
12
Reserved
11
0
Horizontal blank start
Description
31:28
Reserved. Read Only
27:16
Horizontal Blank End
Horizontal blank end, expressed in terms of absolute pixel number relative to the horizontal active display
start
Note: An asserted HBlank will be deasserted when HTotal occurs, regardless of what is programmed in
HBlank end.
15:12
Reserved. Read Only
11:0
Horizontal Blank Start
Horizontal blank start, expressed in terms of absolute pixel number relative to the horizontal active display
start
17.3 HSYNC — Horizontal Sync Register
Address offset:
60008h
Default value:
00000000h
Access:
Read/write
Size:
32 bits
360
®
R
31
28
27
16
Reserved
Horizontal sync end
Bit
15
12
11
Reserved
0
Horizontal sync start
Description
31:28
Reserved. Read Only
27:16
Horizontal Sync End
Horizontal sync end, expressed in terms of absolute pixel number relative to the horizontal active display
start
Note:
1. Minimum HSYNC width is 1 pixel clock.
An asserted HSYNC will be cleared as soon as HTOTAL end is reached, regardless of the value in the
HSYNC End register.
15:12
Reserved. Read Only
11:0
Horizontal Sync Start
Horizontal sync start, expressed in terms of absolute pixel number relative to the horizontal active display
start
Note that when HSYNC start is programmed equal to HBLANK start, both HSYNC and HBLANK will be
asserted on the same pixel clock.
17.4 VTOTAL — Vertical Total Register
Address offset:
6000Ch
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
Reserved
27
16
Vertical total display pixels
Bit
15
11
Reserved
10
0
Vertical active display pixels
Description
31:28
Reserved. Read Only
27:16
Vertical Total Display Pixels
Vertical total display lines. This 12-bit field provides a vertical total up to 4096 lines, encompassing 2048
active display lines, top/bottom border lines, and vertical retrace period.
15:11
Reserved. Read only
10:0
Vertical Active Display Pixels
Vertical active display lines. This 11-bit field provides vertical active display resolution up to 2048 lines.
Note that the first vertical active display line always starts at 0.
361
®
R
17.5 VBLANK — Vertical Blank Register
Address offset:
60010h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
Reserved
27
16
Vertical blank end
Bit
15
12
Reserved
11
0
Vertical blank start
Description
31:28
Reserved. Read Only
27:16
Vertical Blank End
Vertical blank end, expressed in terms of absolute line number relative to the vertical active display start
Note: Vertical blank will be deasserted when vertical total occurs, regardless of what is programmed in
vertical blank end.
15:12
Reserved. Read Only
11:0
Vertical blank start, expressed in terms of absolute line number relative to the vertical active display start
362
®
R
17.6 VSYNC — Vertical Sync Register
Address offset:
60014h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
Reserved
27
16
Vertical sync end
Bit
15
12
Reserved
11
0
Vertical sync start
Description
31:28
Reserved. Read Only
27:16
Vertical Sync End
Vertical sync end, expressed in terms of absolute line numbers relative to the vertical active display start
Notes:
1. When VSYNC start is programmed equal to VBLNK start, both VSYNC and VBLANK will be asserted
on the same pixel clock.
2. VSYNC start programmed beyond the VTOTAL end will prevent the VSYNC start and VSYNC end
from occurring.
15:12
Reserved. Read Only
11:0
Vertical Sync Start
Vertical sync start, expressed in terms of absolute line number relative to the vertical active display start
Notes:
1. Minimum VSYNC width is 2 lines. A VSYNC programmed to 1 scan line does not generate the correct
picture.
2. An asserted VSYNC will be cleared as soon as VTOTAL end is reached, regardless of the value in the
VSYNC End register.
363
®
R
17.7 LCDTV_C — LCD/TV-Out Control Register
Address offset:
60018h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
30
29
28
27
LCD /
TV-out
enable
Reserved
Centering
enable
FP VESA
VGA
mode
24
Reserved
23
16
Reserved
15
14
13
12
11
10
9
8
Reserved
FP / 740
data
ordering
LCD Info.
data
enable
Reserved
VSYNC
control
HSYNC
control
VSYNC
output
control
HSYNC
output
control
7
6
5
4
3
2
1
0
Border
enable
Active
data ½pixel
order
Active
data
polarity
VSYNC
polarity
control
HSYNC
polarity
control
BLANK#
polarity
control
Dot clock
source
Lock dot
clock PLL
N/M regs
Bit
31
Port 2
enable
(rsvd in
GMCH)
Description
LCD / TV-Out Enable
1 = Enable. This bit enables the LCD / TV digital interface. The LCD / TV timing generator is jammed to
pixel 0 of the vertical front porch when this bit is a 0. The timing generator may be ignored, depending
on the LCD timing generator bit (29).
0 = Disable and tristate the whole interface: TVDATA[11:0], BLANK#, TVHSYNC, TVVSYNC, and
TVCLK[1:0]. CLKIN is not disabled and can be used for flat panel hot-plug detection.
30
364
Reserved. Must always be programmed as a zero.
®
R
Bit
29
Description
Centering Enable
0 = Disable. The LCD / TV timing generator controls all display timing when enabled by bit 31 above.
1 = Enable. Centers the VGA active image, as defined in the VGA CRT registers within the LCD/TV active
image.
28
FP VESA VGA Mode
0 = Disable. Use the LCD / TV timing generator. The VGA sync polarity is ignored, and the FP sync
polarity is used. Centering can be enabled for fixed-resolution flat panels or TVs. The Flat Panel Dot
Clock PLL Timing Registers must be used for both flat panels and TVs. After these registers are
written, the Lock Dot Clock PLL N/M Registers must be set to 1, which makes the dot clock PLL use
only the Flat Panel PLL registers.
1 = Enable. Use the VGA timing generator. The VGA sync polarity is passed though, and the FP sync
polarity is ignored. Centering must be disabled. Also sets bit 0 of this register (Lock Dot Clock PLL
N/M Regs) to a 0, which allows normal programming of the Dot Clock PLL registers. This bit should
be disabled when driving a TV.
27:17
Reserved. Must be programmed as 0.
16:15
Reserved
14
FP / 740 Data Ordering
0 = 740-Compliant data ordering:
1 = Flat panel data ordering: R[7:0] ‘ G[7:4] followed by G[3:0] ‘ B[7:0]
13
LCD Information Data Enable. When enabled, it transfers data from the GC to the external device,
during the vertical sync. This transfer should be qualified by blank signal.
0 = Disable
1 = Enable (currently planned for debug purposes)
12
Reserved
11
FPVSYNC Control
1 = FPVSYNC is disabled.
If in FP VESA VGA Mode, then this pin goes to the level of the VGA VSYNC, when disabled.
If not in FP VESA VGA Mode, then this pin goes into the deasserted state, as specified by the VSYNC
0 = FPVSYNC is enabled.
When in FP VESA VGA Mode, then the VGA timing generator is the source of this signal.
When not in FP VESA VGA Mode, then the source of this signal is this timing generator.
10
FPHSYNC Control
1 = FPHSYNC is disabled.
If in FP VESA VGA Mode, then this pin goes to the level of the VGA HSYNC, when disabled.
If not in FP VESA VGA Mode, then this pin goes into the deasserted state, as specified by the HSYNC
0 = FPHSYNC is enabled.
When in FP VESA VGA Mode, then the VGA timing generator is the source of this signal.
When not in FP VESA VGA Mode, then the source of this signal is this timing generator.
365
®
Bit
9
R
Description
FPVSYNC Output Control
1 = Tristates the FPVSYNC pin.
0 = FPVSYNC is active unless LCD / TV-Out Enable is deasserted.
Though this bit is provided, the GC always use VSYNC as output.
8
FPHSYNC Output Control
1 = Tristates the FPHSYNC pin.
0 = FPHSYNC is active unless LCD / TV-Out Enable is deasserted.
7
Border Enable
1 = Border to the LCD / TV encoder is enabled.
0 = Border to the LCD / TV encoder is disabled.
6
Active Data Order
1 = Reversed ½-pixel data ordering: G[3:0] ‘ B[7:0] followed by R[7:0] ‘ G[7:4].
0 = Normal ½-pixel data ordering: R[7:0] ‘ G[7:4] followed by G[3:0] ‘ B[7:0].
5
Active Data Polarity
1 = Inverted pixel data
0 = Normal pixel data
4
VSYNC Polarity Control
1 = Active high
0 = Active low
3
HSYNC Polarity Control
1 = Active HIGH
0 = Active LOW
2
BLANK# Polarity Control
1 = Active HIGH
0 = Active LOW
1
Dot Clock Source
1 = Dot clock PLL reference source is external pin = CLKIN.
0 = Dot clock PLL reference source is the default PLL source.
The CLKIN pin can be used as an interrupt for FP hot plug detection. When the pin is used as a clock, the
interrupt signal is forced to the deassertion level.
The CLKIN / Interrupt pin is always an input; it never is disabled. An internal pull-up is active when the pin
is configured as an interrupt. When configured as a clock, the internal pull-up is disabled.
366
®
R
Bit
Description
Lock Dot Clock PLL N/M Regs
0
1 = Dot Clock PLL N/M registers are locked. = Use the LCD / TV PLL M/N registers and ignore the MSR
register.
0 = Dot Clock PLL N/M registers are writeable. = The MSR register controls which PLL M/N registers are
used.
When supporting either a TV encoder or a flat panel but not in VESA VGA mode, the LCD / TV PLL M/N
registers must be set up for the proper dot clock frequency. Then this bit is written with a 1, which forces
the dot clock PLL to look only at the LCD / TV PLL M/N registers.
17.8 OVRACT — Overlay Active Register
Address offset:
6001Ch
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
31
28
Reserved
27
16
Overlay active end
Bit
15
12
Reserved
11
0
Overlay active start
Description
31:27
Reserved. Read Only
26:16
Overlay Active End
This field takes into account the overlay pipeline delays for turning off the overlay at the end of a scan line.
When LCD / TV is enabled, then the overlay active end is controlled by the LCD / TV-out timing generator
and the bits are used. When LCD / TV is disabled, then the overlay active end is controlled by the VGA
timing generator and uses bits 15:3 for the character clock resolution.
15:12
Reserved. Read Only
11:0
Overlay Active Start
This field takes into account the overlay pipeline delays for lining up X=0 to the first active pixel. When
LCD / TV is enabled, then the overlay active start is controlled by the LCD / TV-out timing generator, and
all the bits are used. When LCD / TV is disabled, then the overlay active start is controlled by the VGA
timing generator and uses bits 15:3 for the character clock.
367
®
R
17.9 BCLRPAT — Border Color Pattern Register
Address offset:
60020h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
A border is sent if Border Enable is on. Also, the same color will be sent during the pseudo-border period in the LCD noscalar mode.
31
0
Reserved
Red
Bit
Description
31:25
Reserved. Read Only
24:16
Red
15:8
Green
7:0
Blue
17.10 Reserved Registers
Address offset:
60024h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
Address offset:
60028h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
Address offset:
60030h
Default value:
00000000h
Access:
Read/Write
Size:
32 bits
368
Green
Blue
®
R
18. Local Memory Interface
18.1 DRT—DRAM Row Type
Address offset :
Default value :
Access :
Size :
03000h
00h
Read/Write
8 bit
This 8-bit register identifies whether or not the local memory is populated.
7
1
Reserved
Bit
7:1
0
0
DRAM
populated
Description
Reserved
DRAM Populated (DP). This bit indicates whether or not the local memory is populated.
0 = No local memory
1 = 4-MB local memory
369
®
R
18.2 DRAMCL—DRAM Control Low
Address offset :
Default value :
Access :
Size :
03001h
17h
Read/Write
8 bit
7
5
Reserved
4
3
2
1
0
Paging
mode
control
RAS-toCAS
override
CAS#
latency
RAS#
riming
RAS#
precharge
timing
Bit
7:5
4
Description
Reserved
Paging Mode Control (PMC)
0 = Page Open Mode. In this mode, the GMCH memory controller tends to leave pages open.
1 = Page Close Mode. In this mode, the GMCH memory controller tends to leave pages closed.
3
RAS-to-CAS Override (RCO).
In units of local memory clock periods (i.e., row activate command to read/write command)
Bit
2
1
0
Determined by CL bit (default)
1
2
CAS# Latency (CL). In units of local memory clock periods.
Bit
CL
0
2
2
1
3
3 (default)
RAS# Act. to Precharge (tRAS) Refresh to RAS# Act. (tRC)
0
5
8
1
7
10 (default)
RAS# Precharge Timing (RPT). This bit controls RAS# precharge (in local memory clocks).
Bit
370
RAS#-to-CAS# delay (tRCD)
RAS# Riming (RT). This bit controls RAS# active to precharge and refresh to RAS# active delay
(in local memory clocks).
Bit
0
RAS#-to-CAS# delay (tRCD)
RAS# Precharge (tRP)
0
2
1
3 (default)
®
R
18.3 DRAMCH—DRAM Control High
Address offset :
Default value :
Access :
Size :
03002h
08h
Read/Write
8 bit
7
5
Reserved
Bit
4
3
DRAM refresh rate
2
0
Reserved
Description
7:5
Reserved
4:3
DRAM Refresh Rate (DRR): The DRAM refresh is controlled using this field. Disabling refresh results in
the eventual loss of DRAM data, although refresh can be briefly disabled without data loss. The field must
be set to normal refresh as soon as possible once DRAM testing is completed.
00 = Refresh disabled
01 = Refresh enabled (default)
11 = Reserved
2:0
Reserved. 0h
371
®
R
372
®
R
19. I/O Control Registers
19.1 HVSYNC—HSYNC/VSYNC Control Register
Address offset: 05000h
Default value:
00000000h
Size:
32 bits
Attribute:
R/W
Bits 19:16 are for DPMS and DDC sync select.
DPMS MODE
HSYNC/VSYNC Control[19:16]
Power On
0000 (i.e., pulse H and V)
Standby
0010 (i.e., pulse V)
Suspend
1000 (i.e., pulse H)
Power Off
1010 (no pulse on H & V)
31
20
Reserved
19
18
17
16
VSYNC
Control
VSYNC
Data
HSYNC
Control
HSYNC
Data
15
1
Reserved
Bit
31:20
19
0
HSYNC/
VSYNC
En
Description
Reserved.
VSYNC Control
Bit 19 (VSYNC Control) and bit 18 (VSYNC Data) are used by the BIOS to take over the sync
during DDC1 communication during POST. The BIOS can force the VSYNC data at the same
time that VSYNC control enables this signal as an output, so that the VSYNC pulse occurs on
every write by the BIOS. This is done to speed up some very slow DDC communications.
0 = Normal VSYNC output
1 = Contents of VSYNC Data will go out to VSYNC pin.
18
VSYNC Data
17
HSYNC Control
0 = Normal HSYNC output
1 = Contents of HSYNC Data will go out to HSYNC pin.
16
HSYNC Data
373
®
R
Bit
Description
Reserved
15:1
HSYNC/VSYNC Enable
0
0 = Hsync and Vsync are deactivated when the internal DAC is disabled (default).
1 = Hsync and Vsync remain active when the internal DAC is disabled through the Module PowerDown
0 Register.
19.2 GPIO Registers
GPIOA
General-Purpose I/O Control Register A
19.2.1
Address offset : 05010h
Default value :
00h, 00h, 000U0000b, 000U0000b
Access :
Read/Write
Size :
32 bit
This register controls the general-purpose I/O pins GPIO0 (DDCSCL pin) and GPIO1 (DDCSDA pin). These two pins
are used specifically to create a Display Data Channel (DDC) serial bus. GPIO0 = DDC Clock (DDCSCL pin); GPIO1 =
DDC data (DDCSDA pin).
31
16
Reserved
15
13
Reserved
7
5
Reserved
374
12
11
10
9
8
GPIO1
data in
GPIO1
data value
GPIO1
data mask
GIO1
direction
value
GPIO1
direction
Mask
4
3
2
1
0
GPIO0
data in
GPIO0
data value
GPIO0
data mask
GPIO0
direction
value
GPIO0
direction
Mask
®
R
Bit
Description
31:16
Reserved
15:13
Reserved
12
GPIO1 Data In (RO)
This is the value that is sampled on the GPIO1 pin as an input.
11
GPIO1 Data Value (R/W)
This value should be placed on the GPIO1 pin as an output. This value is only written into the register if
GPIO1 DATA MASK also is asserted. The value will appear on the pin if this data value is actually
written to this register and the GPIO1 DIRECTION VALUE contains a value that will configure the pin as
an output.
10
GPIO1 Data Mask (R/W)
This mask bit is used to determine whether the GPIO1 DATA VALUE bit should be written into the
register.
9
GPIO1 Direction Value (R/W)
This value should be used to define the output enable of the GPIO1 pin. This value is only written into
the register if GPIO1 DIRECTION MASK also is asserted. The value that will appear on the pin is
defined by what is in the register for the GPIO1 DATA VALUE bit.
8
GPIO1 Direction Mask (R/W)
This mask bit is used to determine whether the GPIO DIRECTION VALUE bit should be written into the
register.
7:5
Reserved
4
GPIO0 Data In (RO)
This value is sampled on the GPIO0 pin as an input.
3
GPIO0 DATA MASK also is asserted. The value will appear on the pin if this data value actually is
written to this register and GPIO0 DIRECTION VALUE contains a value that will configure the pin as an
output.
2
register.
1
375
®
R
Bit
Description
register.
0
GPIOB
General-Purpose I/O Control Register B
19.2.2
Address offset : 05014h
Default value :
00h, 00h, 000U0000b, 000U0000b
Access :
Read/Write
Size :
32 bits
This register controls general-purpose I/O pins GPIO2 (LTVCL pin) and GPIO3 (LTVDA pin). These two pins are used
specifically to create an I2C serial bus interface. GPIO2 = I2C clock (LTVCL pin); GPIO3 = I2C data (LTVDA pin).
31
16
Reserved
15
13
Reserved
7
5
Reserved
Bit
12
11
10
9
8
GPIO3
Data In
GPIO3
Data value
GPIO3
Data mask
GPIO3
Direction
value
GPIO3
Direction
Mask
4
3
2
1
0
GPIO2
Data In
GPIO2
Data value
GPIO2
Data mask
GPIO2
Direction
value
GPIO2
Direction
Mask
Description
31:16
Reserved
15:13
Reserved
12
GPIO3 Data In (RO)
11
GPIO3 DATA MASK also is asserted. The value will appear on the pin if this data value actually is
output.
376
®
R
Bit
10
Description
register.
9
8
register.
7:5
Reserved
4
GPIO2 Data In (RO)
3
GPIO2 DATA MASK also is asserted. The value will appear on the pin if this data value is actually
output.
2
register.
1
0
register.
377
®
R
378
®
R
20. Display and Cursor Registers
The following are cursor, display, and pixel pipe registers in the address range 70000h-7FFFFh.
20.1 DISP_SL—Display Scan Line Count
Memory offset address:
70000h
Default:
0000h
Attributes:
Read Only
This register enables the read-back of the display’s vertical “line counter.” In interlaced display modes, the line counter
is initialized to the field and is incremented by two at each HSYNC.
The display line values are from CRTTG (the CRT timing generator) or the TV/FP timing generator, depending on
whether or not the TV/FP timing generator is enabled and not in the FP VESA VGA mode (LCDTV_C[31]=1 AND
LCDTV_C[28]=0). The values change at the leading edge of HSYNC and can be read safely during the display enable
active time.
In TV/FP centering mode, scan line 0 = the first active scan line of the TV/FP, not the centered active display.
15
12
11
Reserved
Bit
0
Line counter for display [11:0]
Descriptions
15:12
Reserved
11:0
Line counter for display [11:00]
379
®
R
20.2 DISP_SLC—Display Scan Line Count Range Compare
70004h
Default:
0000h
Attributes:
Read Only
The values in the Top and Bottom Line Count Compare registers are compared with the display line values from CRTTG
(the CRT timing generator) or the TV/FP timing generator, depending upon whether or not the TV/FP timing generator
is enabled and not in the FP VESA VGA mode (LCDTV_C[31]=1 AND LCDTV_C[28]=0). The values change at the
leading edge of HSYNC and can be read safely during the display enable active time. The Top-compare register operator
is less than or equal, while the Bottom-compare register operator is greater than or equal. The results of these two
comparisons are communicated to the command stream controller for generating interrupts, status, and command stream
flow control (“wait for scan line”). These registers can be loaded from the command stream and read through the PCI.
In the TV/FP centering mode, scan line 0 = the first active scan line of the TV/FP, not the centered active display.
31
30
In/Ex
28
27
Reserved
15
Top line count compare for display SLC [11:00]
12
11
Reserved
Bit
31
16
0
Bottom line count compare for display SLC [11:00]
Descriptions
Inclusive / Exclusive
1 = Inclusive: Within the range.
0 = Exclusive: Outside of the range.
30:28
Reserved
27:16
Top Line Count Compare for Display SLC [11:00]
This register is used as the top-comparison (less-than-or-equal-to) value with the display vertical “line
counter.”
15:12
Reserved
11:0
Bottom Line Count Compare for Display SLC [11:00]
This register is used as the bottom-comparison (greater-than-or-equal-to) value with the display vertical
“line counter.”
380
®
R
20.3 Pixel Pipeline Control
20.3.1
PIXCONF—Pixel Pipeline Configuration
70008h
Default:
00000000h
Attributes:
Read/Write
31
28
Reserved (0000)
23
21
Reserved (000)
15
14
13
7
27
24
Display
gamma
enable
Overlay
gamma
enable
Reserved
Reserved
19
16
Display color mode
11
10
9
8
Cursor
display
enable
Extended
status
read
CRT
overscan
color
Reserved
Palette
addr
4
3
2
1
0
Reserved
(0)
Reserved
(system
write 32)
Reserved
(0)
VGA
wrap
GUI
mode
Bit
31:28
25
12
5
Reserved (000)
26
CRT
control
Reserved
8-bit
DAC
enable
20
27
Descriptions
Reserved (0000)
Display Path (Graphics) Gamma Enable (see note)
0 = 16- and 24-bpp graphics data bypasses palette (default).
1 = 16- and 24-bpp graphics data goes through palette.
26
Overlay Path Gamma Enable (see note)
0 = Video data bypasses palette (default).
1 = Video data goes through palette. This is useful when alpha-blending the overlay with the primary
display in order to provide gamma correction for the display device. The overlay gamma correction
should be set up to un-gamma the overlay surface, bringing it into the linear space before performing
the alpha blending. Both the primary display (27 = 1) and the overlay (26=1) should be passed through
the palette after alpha-blending, in order to provide proper gamma correction for the display device.
381
®
Bit
25:21
20
R
Descriptions
Reserved (0s)
CRT Control Signal Delay
0 = CRT Display Enable and CRT Blank are delayed for standard VGA compatibility (default).
1 = CRT Display Enable and CRT Blank are not delayed.
This bit affects the CRT Display Enable and CRT Blank signal delay with respect to CRT HSYNC and CRT
VSYNC, when the standard VGA pixel pipeline is used by CRT display engine.
This bit has no effect on the flat panel centering or optimized timing modes.
19:16
Display Color Mode
0000 = CRT standard VGA text and graphics mode and 1-bit/2-bit/4-bit packed graphics mode (default)
0001 = Reserved
0010 = CRT 8-bit packed extended graphics mode
0011 = Reserved
0100 = CRT 16-bit packed (5-5-5) extended graphics mode (Targa compatible)
0101 = CRT 16-bit packed (5-6-5) extended graphics mode (XGA compatible)
0110 = CRT 24-bit extended graphics mode compressed
0111 = CRT 24-bit extended graphics mode uncompressed. In this mode, pixels are stored only in the
lower three bytes (plane 0,1,2) of each double word, and the most significant byte of each double
word (plane 3) is not used.
15
8-Bit DAC Enable
0 = 6-bit DAC (default)
1 = 8-bit DAC
14:13
12
Reserved
Hardware Cursor Display Enable
1 = Enable
Software should always set this bit to 1. The setting of this bit to 1 should not cause harm, even in the VGA
mode.
11
Enable Extended Status Read Mode
1 = The enabling of this bit makes available the status of the internal state machines and the values of the
red and green data in the input holding register through the normal DAC register ports. The register
ports are redefined as follows when this bit is set:
DACMASK = Returns the red input data holding value.
DACWX = Returns the green input data holding value.
DACSTATE = Returns the status of the internal state machines in bits [7:2].
10
CRT Overscan Color
1 = Enable protected CRT overscan color (overscan[0])
9
382
Reserved
®
R
Bit
8
Descriptions
Palette Addressing
1 = Enable extended palette addressing (Enables access to all 8 locations.)
7:5
4
Reserved (000)
Reserved (0) (Extended text - Forced to 0)
0 = Standard VGA text (default)
1 = Extended VGA text with fonts rearranged for maximum page mode hit
3
Reserved (0) (System Write 32)
0 = System writes to display memory may be assembled for 64-bit cycles (default).
1 = System writes to display memory go as 32-bit cycles.
2
Reserved (0)
1
VGA Wrap
0 = 256-KB wrap state (default) for memory starting at A0000
1 = Don’t wrap
0
GUI Mode
0 = Standard VGA and extended 4-bpp, 16-color resolutions (default). Can still access memory in linear
mode.
1 = High resolution (i.e., not VGA or extended planar)
Transition from VGA modes to hi-res mode or the opposite:
Before writing to PIXCONF[0] and turning the display on, the software will turn off the display engine (i.e.,
screen off) using SR01[Screen Off] and will wait for at least a couple of HSYNC periods and no more than
a couple of VSYNC periods. (Since one of the isochronous streams is DRAM refresh (controlled by
DRAMCXH[DRAM Refresh Status]), the wait should not be so long as to cause the DRAM content to
degrade.) This should ensure that all the data requested from the display engine will be out of the local
memory interface before PIXCONF is touched. In addition, while switching from hi-res to VGA or VGA to hires, the software will ensure that all the other isochronous streams are off before programming
PIXCONF[0].
Note:
Bits [27:24] are not normally used by the graphics BIOS or the drivers because the gamma values are specific to a
particular display device and apply to two color or hi-color modes (16 and 24 bit). It is necessary to program the
palette first with the gamma-adjusted values. There is only one palette, so if both 3:2 are set, they have the same
gamma adjustments. Typical code and typical drivers leave these bits set at zero.
383
®
R
20.4 BLTCNTL—BLT Control
7000Ch
Default:
0000h
Attributes:
Read/Write
15
8
Reserved
7
6
5
Reserved
4
Default color
expansion mode
Bit
3
2
Reserved
Descriptions
15:6
Reserved (0s)
5:4
Default Color Expansion Mode—R/W
00 = 8 bits per pixel (default)
01 = 16 bits per pixel
10 = 24 bits per pixel
11 = Reserved
3:1
0
Reserved (000)
BLT Status—RO
This read-only bit reflects the busy status of the BLT Engine only.
0 = Idle (default)
1 = Busy
384
1
0
Reserved
BLT
status
®
R
20.5 Reserved Register
70010h
Default:
0000h
Attributes:
Read/Write
15
8
Reserved = 00h
7
0
Reserved = 00h
20.6 SWF[1:3]—Software Flag Registers
SWF1 = 70014h
SWF2 = 70018h
SWF3 = 7001Ch
Default:
00000000h
Attributes:
Read/Write
These 32-bit registers are used as scratch pad space in the BIOS and have no effect on hardware.
385
®
20.6.1
R
DPLYBASE—Display Base Address Register
70020h
Default:
0000h
Attributes:
Read/Write
The display can be read from graphics memory. This register is the display staging register when written. The load
register becomes the active register at the asserting edge of the vertical sync. The read-back register is from the active
register.
31
26
25
3
Reserved
Bit
Reserved
25:3
Display Base Address Bits [25:03]
This is the base address of the display.
2:0
Reserved
386
0
Reserved
Descriptions
31:26
2
®
R
20.6.2
DPLYSTAS—Display Status Select Register
70024h
Default:
0000h
Attributes:
Read/Write
This register selects the proper events to be signaled to the Interrupt Control Register in the command stream. Status bits
0 and 1 are logically ORed to become the Vertical Blank status bit, and status bits 8, 9, and 10 are logically ORed to
become the Display Event status bit. These two sets of status bits are on separate bytes for future expansion purposes.
The architecture allows each byte to be logically ORed. The enable for each status bit is in the corresponding bit or the
upper word. If a status enable bit is asserted, then the corresponding status bit is considered in the interrupt generation.
The status bits capture the event if enabled and are cleared by writing a 1 to the bit. This register has separate byte
enables for writing.
31
27
Reserved
26
25
24
FP hot
plug
enable
VSYNC
enable
Vert line
comp
enable
17
16
Vert
blank en
OVL reg
ppd en
10
9
8
FP hot
plug
interrupt
VSYNC
Vert line
comp
1
0
Vert blnk
dply
event sel
OVL reg
upd
Vblnk sel
23
18
Reserved
15
14
FP hot
plug
status
11
Reserved
7
2
Reserved
Bit
31:27
26
Descriptions
Reserved
Flat Panel Hot Plug Detect Enable
0 = Flat panel hot plug detect disabled
1 = Flat panel hot plug detect enabled
25
Vertical Sync Status Enable
0 = Vertical sync status disabled
1 = Vertical sync status enabled
387
®
Bit
24
R
Descriptions
Display Line Compare Enable
0 = Display line compare status disabled
1 = Display line compare status enabled
23:18
17
Reserved
Vertical Blank Enable
0 = Vertical blank status disabled
1 = Vertical blank status enabled
16
Overlay Registers Updated Enable
0 = Overlay registers have been updated while vertical blank status disabled.
1 = Overlay registers have been updated while vertical blank status enabled.
15
Flat Panel Hot Plug Detect Status
This bit is the state of the TVCLKIN pin of the TV/flat panel interface. This pin signals an interrupt when it is
LOW. When an interrupt is asserted on the pin, this status bit reads back as a 1. This bit is forced low when
the TVCLKIN pin is selected as a clock input reference to the dot clock PLL and NOT an Interrupt pin.
0 = Flat Panel Hot Plug Detect asserted
1 = Flat Panel Hot Plug Detect not asserted (forced to 1 when used as CLKIN)
14:11
10
Reserved
Flat Panel Hot Plug Detect (edge catcher)
This bit is the edge detector for bit 15 above. It is set to 1 when it detects a low-to-high transition.
0 = Flat Panel Hot Plug Detect not asserted. Bit 15 has not transitioned from 1 to 0 (forced to 0 when used
as CLKIN).
1 = Flat Panel Hot Plug Detect asserted. Bit 15 has transitioned from 1 to 0.
9
Vertical Sync Status
0 = Vertical Sync not asserted
1 = Vertical Sync asserted
8
Display Line Compare Status
0 = Display Line Compare Status not asserted
1 = Display Line Compare Status asserted
7:2
1
Reserved
Vertical Blank Status
0 = Vertical Blank Status not asserted
1 = Vertical Blank Status asserted
0
Overlay Registers Updated Status
0 = Overlay registers have been updated while Vertical Blank Status is not asserted.
1 = Overlay registers have been updated while Vertical Blank Status is asserted.
388
®
R
20.7 Hardware Cursor
The hardware cursor registers are memory mapped and accessible through 32-bit accesses.
The hardware cursor uses colors from extended palette index 4 through 7.
20.7.1
CURCNTR—Cursor Control Register
70080h
Default:
0000h
Attributes:
Read/Write
This register is double-buffered. The load register is transferred into the active register on the asserting edge of Vertical
Sync.
15
5
Reserved
Bit
15:5
4
4
3
Coord.
origin
select
Res
2
0
Cursor mode
select
Descriptions
Reserved
Cursor Coordinate System Origin Select
0 = Selects the outermost upper-left-hand corner of the screen border as the origin for the coordinate
system used to position the cursor (default).
1 = Selects the upper-left-hand corner of the active display area as the origin for the coordinate system
used to position the cursor.
3
2:0
Reserved
Cursor Mode Select
These three bits select the mode for the cursor, as follows:
000 = Cursor is disabled. This is the default after reset.
001 = 32x32, 2-bpp, AND/XOR, 2-plane mode
010 = Reserved
011 = Reserved
100 = 64x64, 2-bpp, 3-color and transparency mode
101 = 64x64, 2-bpp, AND/XOR, 2-plane mode
110 = 64x64, 2-bpp, 4-color mode
111 = Reserved
389
®
20.7.2
R
CURBASE—Cursor Base Address Register
70084h
Default:
0000h
Attributes:
Read/Write
The cursor can be read only from system memory. This register is double-buffered. The load register becomes the active
register at the asserting edge of Vertical Sync.
31
29
Reserved
Bit
28
8
Cursor base address bits [28:08]
7
0
Reserved
Descriptions
31:8
Cursor Base Address Bits [28:08]
These 21 bits provide the most-significant bits of a 29-bit address of the graphics non-cacheable system
memory space where the 512-byte to 1-KB cursor data space for the cursor is to be located.
7:0
Reserved
390
®
R
20.7.3
CURPOS—Cursor Position Register
70088h
Default:
0000h
Attributes:
Read/Write
CURPOS is double-buffered. The load register becomes the active register at the asserting edge of Vertical Sync.
31
30
Xposition
sign bit
Bit
31
26
Reserved (0000)
Yposition
sign bit
15
27
14
Cursor Y-position magnitude bits [10:0]
11
Reserved (0000)
16
10
0
Cursor X-position magnitude bits [10:0]
Descriptions
Cursor Y-Position Sign Bit
This bit provides the sign bit of a signed 12-bit value that specifies the horizontal position of the cursor.
(Default: 0)
30:27
Reserved
26:16
Cursor Y-Position Magnitude Bits 10-0
This field provides the magnitude bits of a signed 12-bit value that specifies the vertical position of the
cursor. The sign bit of this value is provided by bit 31of this register. (Default: 0)
15
Cursor X-Position Sign Bit
This bit provides the sign bit of the signed 12-bit value that specifies the horizontal position of the cursor.
(Default: 0)
14:11
Reserved
10:0
Cursor X-Position Magnitude Bits 10-0
These 11 bits provide the signed 12-bit value that specifies the horizontal position of the cursor. The sign
bit is provided by bit 15 of this register. (Default: 0)
391
®
R
392
R
®
21. Mode Parameters
This chapter contains the register programming information on a per-mode basis.
Refer to the appropriate table for the specific values to use in order to correctly program the graphics adapter for the
desired mode and frequency combination. Programming the graphics adapter with values not included in these tables
may damage the adapter and any connected output devices.
393
®
R
Parameters for Screen Resolution / Refresh Rate:
320x200_70Hz =
Dot clock value
25,
M value
0x0013,
// M
N value
0x0003,
// N
P value
0x50,
// P
CR00
0x2E,
CR01
0x27,
CR02
0x28,
CR03
0x90,
CR04
0x2A,
CR05
0x90,
CR06
0xBF,
CR07
0x1F,
CR09
0xC0,
CR10
0x9C,
CR11
0x0E,
CR12
0x8F,
CR13
0x28,
CR15
0x96,
CR16
0xB9,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
Watermark for 8 bpp at 100-MHz local memory speed
0x0070C000,
0x0020C000,
0x0020C000,
394
// 25.175-MHz dot clock
®
R
0x0070C000,
0x0020C000,
0x0020C000,
MSR mask
0x40
395
®
R
320x240_70Hz =
Dot clock value
31,
M value
0x0013,
//M
N value
0x0002,
//N
P value
0x50,
//P
CR00
0x31,
CR01
0x27,
CR02
0x28,
CR03
0x92,
CR04
0x2C,
CR05
0x90,
CR06
0x05,
CR07
0x3E,
CR09
0xC0,
CR10
0xE9,
CR11
0x0C,
CR12
0xDF,
CR13
0x28,
CR15
0xE9,
CR16
0xFC,
CR30
0x02,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x0070C000,
0x0020C000,
0x0040A000,
396
//31-MHz dot clock
®
R
0x0070C000,
0x0020C000,
0x0040A000,
MSR mask
0xC0
397
®
R
352x480_70Hz =
Dot clock value
15,
M value
0x000D,
//M
N value
0x0001,
//N
P value
0x50,
//P
CR00
0x33,
CR01
0x2B,
CR02
0x2B,
CR03
0x97,
CR04
0x2D,
CR05
0x91,
CR06
0xf2,
CR07
0x10,
CR09
0x40,
CR10
0xe0,
CR11
0x03,
CR12
0xdF,
CR13
0x2C,
CR15
0xdF,
CR16
0xf3,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x0070C000,
0x0020C000,
0x00408000,
398
//15.68-MHz dot clock
®
R
0x0070C000,
0x0020C000,
0x00408000,
MSR mask
0xC0
399
®
R
352x576_70Hz =
Dot clock value
19,
M value
0x0011,
//M
N value
0x0004,
//N
P value
0x40,
//P
CR00
0x35,
CR01
0x2B,
CR02
0x2B,
CR03
0x99,
CR04
0x2D,
CR05
0x92,
CR06
0x56,
CR07
0x10,
CR09
0x40,
CR10
0x40,
CR11
0x03,
CR12
0x3F,
CR13
0x2C,
CR15
0x3F,
CR16
0x57,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x0070C000,
0x0020C000,
0x00408000,
400
//19-MHz dot clock
®
R
0x0070C000,
0x0020C000,
0x00408000,
MSR mask
0xC0
401
®
R
400x300_70Hz =
Dot clock value
49,
M value
0x001F,
//M
N value
0x0006,
//N
P value
0x40,
//P
CR00
0x3F,
CR01
0x31,
CR02
0x32,
CR03
0x80,
CR04
0x38,
CR05
0x1D,
CR06
0x86,
CR07
0xF0,
CR09
0xE0,
CR10
0x63,
CR11
0x06,
CR12
0x57,
CR13
0x32,
CR15
0x63,
CR16
0x7B,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x0070C000,
0x00409000,
0x00409000,
402
//49-MHz dot clock
®
R
0x0070C000,
0x00409000,
0x00409000,
MSR mask
0x00
403
®
R
512x384_70Hz =
Dot clock value
82,
M value
0x0027,
//M
N value
0x000A,
//N
P value
0x30,
//P
CR00
0x53,
CR01
0x3F,
CR02
0x40,
CR03
0x94,
CR04
0x47,
CR05
0x0E,
CR06
0x3B,
CR07
0xFD,
CR09
0xE0,
CR10
0x0E,
CR11
0x01,
CR12
0xFF,
CR13
0x40,
CR15
0x0E,
CR16
0x2D,
CR30
0x03,
CR31
0x02,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x00,
0x0040A000,
0x0040A000,
0x0040A000,
404
//82-MHz dot clock
®
R
0x0040A000,
0x0040A000,
0x0040A000,
MSR mask
0xC0
405
®
R
640x350_85Hz =
Dot clock value
31,
M value
0x0013,
//M
N value
0x0002,
//N
P value
0x40,
//P
CR00
0x63,
CR01
0x4F,
CR02
0x4F,
CR03
0x87,
CR04
0x53,
CR05
0x9B,
CR06
0xBB,
CR07
0x10,
CR09
0x40,
CR10
0x7D,
CR11
0x00,
CR12
0x5D,
CR13
0x50,
CR15
0x5D,
CR16
0xBC,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000,
0x22005000,
0x22007000,
406
®
R
0x22003000,
0x22005000,
0x22007000,
MSR mask
0x80
407
®
R
640x400_70Hz =
Dot clock value
25,
M value
0x0013,
//M
N value
0x0003,
//N
P value
0x40,
//P
CR00
0x5F,
CR01
0x4F,
CR02
0x50,
CR03
0x82,
CR04
0x54,
CR05
0x80,
CR06
0xBF,
CR07
0x1F,
CR09
0x40,
CR10
0x9C,
CR11
0x0E,
CR12
0x8F,
CR13
0x50,
CR15
0x96,
CR16
0xB9,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x00409000, //40A000
0x0040A000,
0x0040A000,
408
®
R
0x00409000, //40A000
0x0040A000,
0x0040A000,
MSR mask
0x40
409
®
R
640x400_85Hz =
Dot clock value
31,
M value
0x0013,
//M
N value
0x0002,
//N
P value
0x40,
//P
CR00
0x63,
CR01
0x4F,
CR02
0x4F,
CR03
0x87,
CR04
0x53,
CR05
0x9B,
CR06
0xBB,
CR07
0x10,
CR09
0x40,
CR10
0x90,
CR11
0x03,
CR12
0x8F,
CR13
0x50,
CR15
0x8F,
CR16
0xBC,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000,
0x22005000,
0x22007000,
410
®
R
0x22003000,
0x22005000,
0x22007000,
MSR mask
0x40
411
®
R
640x480_60Hz =
Dot clock value
25,
M value
0x0013,
//M
N value
0x0003,
//N
P value
0x40,
//P
CR00
0x5F,
CR01
0x4F,
CR02
0x50,
CR03
0x82,
CR04
0x51,
CR05
0x9D,
CR06
0x0B,
CR07
0x10,
CR09
0x40,
CR10
0xE9,
CR11
0x0B,
CR12
0xDF,
CR13
0x50,
CR15
0xE7,
CR16
0x04,
CR30
0x02,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22002000
0x22004000
0x22006000,
412
®
R
0x22002000
0x22004000
0x22006000
MSR mask
0xC0
413
®
R
640x480_70Hz =
Dot clock value
28,
M value
0x0053,
//M
N value
0x0010,
//N
P value
0x40,
//P
CR00
0x61,
CR01
0x4F,
CR02
0x4F,
CR03
0x85,
CR04
0x52,
CR05
0x9A,
CR06
0xF2,
CR07
0x10,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xDF,
CR13
0x50,
CR15
0xDF,
CR16
0xF3,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22002000,
0x22004000,
0x22005000,
414
//28-MHz dot clock
®
R
0x22002000,
0x22004000,
0x22005000,
MSR mask
0xC0
415
®
R
640x480_72Hz =
Dot clock value
31,
M value
0x0013,
//M
N value
0x0002,
//N
P value
0x40,
//P
CR00
0x63,
CR01
0x4F,
CR02
0x4F,
CR03
0x87,
CR04
0x52,
CR05
0x97,
CR06
0x06,
CR07
0x0F,
CR09
0x40,
CR10
0xE8,
CR11
0x0B,
CR12
0xDF,
CR13
0x50,
CR15
0xDF,
CR16
0x07,
CR30
0x02,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000
0x22005000
0x22007000
416
®
R
0x22003000
0x22005000
0x22007000
MSR mask
0xC0
417
®
R
640x480_75Hz =
Dot clock value
31,
M value
0x0013,
//M
N value
0x0002,
//N
P value
0x40,
//P
CR00
0x64,
CR01
0x4F,
CR02
0x4F,
CR03
0x88,
CR04
0x51,
CR05
0x99,
CR06
0xF2,
CR07
0x10,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xDF,
CR13
0x50,
CR15
0xDF,
CR16
0xF3,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000
0x22005000
0x22007000
418
®
R
0x22003000
0x22005000
0x22007000
MSR mask
0xC0
419
®
R
640x480_85Hz =
Dot clock value
36,
M value
0x0010,
//M
N value
0x0001,
//N
P value
0x40,
//P
CR00
0x63,
CR01
0x4F,
CR02
0x4F,
CR03
0x87,
CR04
0x56,
CR05
0x9D,
CR06
0xFB,
CR07
0x10,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xDF,
CR13
0x50,
CR15
0xDF,
CR16
0xFC,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000
0x22005000
0x22107000
420
//36-MHz dot clock
®
R
0x22003000
0x22005000
0x22107000
MSR mask
0xC0
421
®
R
720x400_85Hz =
Dot clock value
35,
M value
0x0045,
//M
N value
0x000A,
//N
P value
0x40,
//P
CR00
0x70,
CR01
0x59,
CR02
0x59,
CR03
0x94,
CR04
0x5D,
CR05
0x86,
CR06
0xBC,
CR07
0x1F,
CR09
0x40,
CR10
0x90,
CR11
0x03,
CR12
0x8F,
CR13
0x5A,
CR15
0x8F,
CR16
0xBD,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x01,
0x22003000
0x22005000,
0x22107000,
422
®
R
0x22003000
0x22005000,
0x22107000,
MSR mask
0x40
423
®
R
720x480_60Hz =
Dot clock value
28,
M value
0x0053,
//M
N value
0x0010,
//N
P value
0x40,
//P
CR00
0x6B,
CR01
0x59,
CR02
0x59,
CR03
0x8F,
CR04
0x5B,
CR05
0x84,
CR06
0xEF,
CR07
0x10,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xE0,
CR13
0x5A,
CR15
0xDF,
CR16
0xF0,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x22002000,
0x22004000,
0x22006000,
424
®
R
0x22002000,
0x22004000,
0x22006000,
MSR mask
0x00
425
®
R
720x480_75Hz =
Dot clock value
35,
M value
0x0021,
//M
N value
0x0004,
//N
P value
0x40,
//P
CR00
0x6F,
CR01
0x59,
CR02
0x5A,
CR03
0x92,
CR04
0x65,
CR05
0x8E,
CR06
0xF4,
CR07
0x1F,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xDF,
CR13
0x5A,
CR15
0xE0,
CR16
0xF4,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x22003000,
0x22004000,
0x22008000,
426
//35-MHz dot clock
®
R
0x22003000,
0x22004000,
0x22008000,
MSR mask
0x00
427
®
R
720x480_85Hz =
Dot clock value
40,
M value
0x0008,
//M
N value
0x0001,
//N
P value
0x30,
//P
CR00
0x6F,
CR01
0x59,
CR02
0x5A,
CR03
0x92,
CR04
0x65,
CR05
0x8E,
CR06
0xF7,
CR07
0x1F,
CR09
0x40,
CR10
0xE0,
CR11
0x03,
CR12
0xDF,
CR13
0x5A,
CR15
0xE0,
CR16
0xF7,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x22003000,
0x22004000,
0x22008000,
428
//40-MHz dot clock
®
R
0x22003000,
0x22004000,
0x22008000,
MSR mask
0x00
429
®
R
720x576_60Hz =
Dot clock value
33,
M value
0x0014,
//M
N value
0x0002,
//N
P value
0x40,
//P
CR00
0x6D,
CR01
0x59,
CR02
0x59,
CR03
0x91,
CR04
0x5C,
CR05
0x85,
CR06
0x53,
CR07
0x10,
CR09
0x40,
CR10
0x40,
CR11
0x03,
CR12
0x3F,
CR13
0x5A,
CR15
0x3F,
CR16
0x54,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x01,
0x22002000,
0x22005000,
0x22006000,
430
//33-MHz dot clock
®
R
0x22002000,
0x22005000,
0x22006000,
MSR mask
0x00
431
®
R
720x576_75Hz =
Dot clock value
43,
M value
0x0029,
//M
N value
0x000A,
//N
P value
0x30,
//P
CR00
0x71,
CR01
0x59,
CR02
0x5A,
CR03
0x94,
CR04
0x65,
CR05
0x8E,
CR06
0x58,
CR07
0xF0,
CR09
0x60,
CR10
0x40,
CR11
0x03,
CR12
0x3F,
CR13
0x5A,
CR15
0x40,
CR16
0x58,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000,
0x22006000,
0x22008000,
432
//43-MHz dot clock
®
R
0x22004000,
0x22006000,
0x22008000,
MSR mask
0x00
433
®
R
720x576_85Hz =
Dot clock value
49,
M value
0x001F,
//M
N value
0x0006,
//N
P value
0x30,
//P
CR00
0x71,
CR01
0x59,
CR02
0x5A,
CR03
0x94,
CR04
0x65,
CR05
0x8E,
CR06
0x5B,
CR07
0xF0,
CR09
0x60,
CR10
0x40,
CR11
0x03,
CR12
0x3F,
CR13
0x5A,
CR15
0x40,
CR16
0x5B,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000,
0x22006000,
0x22009000,
434
®
R
0x22004000,
0x22006000,
0x22009000,
MSR mask
0x0
435
®
R
800x600_56Hz =
Dot clock value
36,
M value
0x0010,
//M
N value
0x0001,
//N
P value
0x40,
//P
CR00
0x7B,
CR01
0x63,
CR02
0x63,
CR03
0x9F,
CR04
0x66,
CR05
0x8F,
CR06
0x6F,
CR07
0x10,
CR09
0x40,
CR10
0x58,
CR11
0x0A,
CR12
0x57,
CR13
0xC8,
CR15
0x57,
CR16
0x70,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x01,
0x22003000
0x22005000
0x22107000
436
//36-MHz dot clock
®
R
0x22003000
0x22005000
0x22107000
MSR mask
0x00
437
®
R
800x600_60Hz =
Dot clock value
40,
M value
0x0008,
//M
N value
0x0001,
//N
P value
0x30,
//P
CR00
0x7F,
CR01
0x63,
CR02
0x63,
CR03
0x83,
CR04
0x68,
CR05
0x18,
CR06
0x72,
CR07
0x10,
CR09
0x40,
CR10
0x58,
CR11
0x0C,
CR12
0x57,
CR13
0xC8,
CR15
0x57,
CR16
0x73,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22003000
0x22006000
0x22108000
438
//40-MHz dot clock
®
R
0x22003000
0x22006000
0x22108000
MSR mask
0x00
439
®
R
800x600_70Hz =
Dot clock value
45,
M value
0x0054,
//M
N value
0x0015,
//N
P value
0x30,
//P
CR00
0x7D,
CR01
0x63,
CR02
0x63,
CR03
0x81,
CR04
0x68,
CR05
0x12,
CR06
0x6f,
CR07
0x10,
CR09
0x40,
CR10
0x58,
CR11
0x0b,
CR12
0x57,
CR13
0x64,
CR15
0x57,
CR16
0x70,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000,
0x22007000,
0x2210A000,
440
//45-MHz dot clock
®
R
0x22004000,
0x22007000,
0x2210A000,
MSR mask
0x00
441
®
R
800x600_72Hz =
Dot clock value
50,
M value
0x0017,
//M
N value
0x0004,
//N
P value
0x30,
//P
CR00
0x7D,
CR01
0x63,
CR02
0x63,
CR03
0x81,
CR04
0x6A,
CR05
0x19,
CR06
0x98,
CR07
0x10,
CR09
0x40,
CR10
0x7C,
CR11
0x02,
CR12
0x57,
CR13
0xC8,
CR15
0x57,
CR16
0x99,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000
0x22007000
0x2210A000
442
//50-MHz dot clock
®
R
0x22004000
0x22007000
0x2210A000
MSR mask
0x00
443
®
R
800x600_75Hz =
Dot clock value
49,
M value
0x001F,
//M
N value
0x0006,
//N
P value
0x30,
//P
CR00
0x7F,
CR01
0x63,
CR02
0x63,
CR03
0x83,
CR04
0x65,
CR05
0x0F,
CR06
0x6F,
CR07
0x10,
CR09
0x40,
CR10
0x58,
CR11
0x0B,
CR12
0x57,
CR13
0xC8,
CR15
0x57,
CR16
0x70,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000
0x22007000
0x2210B000
444
®
R
0x22004000
0x22007000
0x2210B000
MSR mask
0x00
445
®
R
800x600_85Hz =
Dot clock value
56,
M value
0x0049,
//M
N value
0x000E,
//N
P value
0x30,
//P
CR00
0x7E,
CR01
0x63,
CR02
0x63,
CR03
0x82,
CR04
0x67,
CR05
0x0F,
CR06
0x75,
CR07
0x10,
CR09
0x40,
CR10
0x58,
CR11
0x0B,
CR12
0x57,
CR13
0xC8,
CR15
0x57,
CR16
0x76,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22004000
0x22108000
0x2210b000
446
®
R
0x22004000
0x22108000
0x2210b000
MSR mask
0x00
447
®
R
854x480_60Hz =
Dot clock value
43,
M value
0x0029,
//M
N value
0x000a,
//N
P value
0x30,
//P
CR00
0x80,
CR01
0x6a,
CR02
0x6b,
CR03
0x83,
CR04
0x6f,
CR05
0x1a,
CR06
0xef,
CR07
0x1f,
CR09
0x40,
CR10
0xe0,
CR11
0x03,
CR12
0xdf,
CR13
0x6b,
CR15
0xe0,
CR16
0xef,
CR30
0x01,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x0
0x0,
0x0,
448
//43-MHz dot clock
®
R
0x0
0x0,
0x0,
MSR mask
0x00
449
®
R
854x480_75Hz =
Dot clock value
41,
M value
0x002b,
//M
N value
0x000b,
//N
P value
0x30,
//P
CR00
0x84,
CR01
0x6a,
CR02
0x6b,
CR03
0x87,
CR04
0x71,
CR05
0x1c,
CR06
0xf4,
CR07
0x1f,
CR09
0x40,
CR10
0xe0,
CR11
0x03,
CR12
0xdf,
CR13
0x6b,
CR15
0xe0,
CR16
0xf4,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x0
0x0,
0x0,
450
®
R
0x0
0x0,
0x0,
MSR mask
0x00
451
®
R
854x480_85Hz =
Dot clock value
48,
M value
0x000a,
//M
N value
0x0001,
//N
P value
0x30,
//P
CR00
0x86,
CR01
0x6a,
CR02
0x6b,
CR03
0x89,
CR04
0x72,
CR05
0x1d,
CR06
0xf7,
CR07
0x1f,
CR09
0x40,
CR10
0xe0,
CR11
0x03,
CR12
0xdf,
CR13
0x6b,
CR15
0xe0,
CR16
0xf7,
CR30
0x01,
CR31
0x01,
CR32
0x01,
CR33
0x01,
CR35
0x00,
CR39
0x00,
0x0
0x0,
0x0,
452
//Dot clock
®
R
0x0
0x0,
0x0,
MSR mask
0x00
453
®
R
1024x768_60Hz =
Dot clock value
65,
M value
0x003F,
//M
N value
0x000A,
//N
P value
0x30,
//P
CR00
0xA3,
CR01
0x7F,
CR02
0x7F,
CR03
0x87,
CR04
0x82,
CR05
0x93,
CR06
0x24,
CR07
0x10,
CR09
0x40,
CR10
0x02,
CR11
0x08,
CR12
0xFF,
CR13
0x80,
CR15
0xFF,
CR16
0x25,
CR30
0x03,
CR31
0x02,
CR32
0x03,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22005000
0x22109000
0x2220D000
454
//65-MHz dot clock
®
R
0x22005000
0x22109000
0x2220D000
MSR mask
0xC0
455
®
R
1024x768_70Hz =
Dot clock value
75,
M value
0x0017,
//M
N value
0x0002,
//N
P value
0x30,
//P
CR00
0xA1,
CR01
0x7F,
CR02
0x7F,
CR03
0x85,
CR04
0x82,
CR05
0x93,
CR06
0x24,
CR07
0x10,
CR09
0x40,
CR10
0x02,
CR11
0x08,
CR12
0xFF,
CR13
0x80,
CR15
0xFF,
CR16
0x25,
CR30
0x03,
CR31
0x02,
CR32
0x03,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22005000
0x2210A000
0x2220F000
456
//75-MHz dot clock
®
R
0x22005000
0x2210A000
0x2220F000
MSR mask
0xC0
457
®
R
1024x768_75Hz =
Dot clock value
78,
M value
0x0050,
//M
N value
0x0017,
//N
P value
0x20,
//P
CR00
0x9F,
CR01
0x7F,
CR02
0x7F,
CR03
0x83,
CR04
0x81,
CR05
0x8D,
CR06
0x1E,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0x80,
CR15
0xFF,
CR16
0x1F,
CR30
0x03,
CR31
0x02,
CR32
0x03,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22006000
0x2210B000
0x22210000
458
®
R
0x22006000
0x2210B000
0x22210000
MSR mask
0x00
459
®
R
1024x768_85Hz =
Dot clock value
94,
M value
0x003D,
//M
N value
0x000E,
//N
P value
0x20,
//P
CR00
0xA7,
CR01
0x7F,
CR02
0x7F,
CR03
0x8B,
CR04
0x85,
CR05
0x91,
CR06
0x26,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0x80,
CR15
0xFF,
CR16
0x27,
CR30
0x03,
CR31
0x02,
CR32
0x03,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x22007000
0x2220E000
0x22212000
460
®
R
0x22007000
0x2220E000
0x22212000
MSR mask
0x00
461
®
R
1152x864_60Hz =
Dot clock value
80,
M value
0x0008,
//M
N value
0x0001,
//N
P value
0x20,
//P
CR00
0xB3,
CR01
0x8F,
CR02
0x8F,
CR03
0x97,
CR04
0x93,
CR05
0x9f,
CR06
0x87,
CR07
0x10,
CR09
0x40,
CR10
0x60,
CR11
0x03,
CR12
0x5F,
CR13
0x90,
CR15
0x5f,
CR16
0x88,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x00,
0x2220C000,
0x22210000,
0x22415000,
462
//80-MHz dot clock
®
R
0x2220C000,
0x22210000,
0x22415000,
MSR mask
0x00
463
®
R
1152x864_70Hz =
Dot clock value
96,
M value
0x000a,
//M
N value
0x0001,
//N
P value
0x20,
//P
CR00
0xbb,
CR01
0x8F,
CR02
0x8F,
CR03
0x9f,
CR04
0x98,
CR05
0x87,
CR06
0x82,
CR07
0x10,
CR09
0x40,
CR10
0x60,
CR11
0x03,
CR12
0x5F,
CR13
0x90,
CR15
0x5F,
CR16
0x83,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x00,
0x22107000,
0x22210000,
0x22415000
464
//96-MHz dot clock
®
R
0x22107000,
0x22210000,
0x22415000
MSR mask
0x00
465
®
R
1152x864_72Hz =
Dot clock value
99,
M value
0x001f,
//M
N value
0x0006,
//N
P value
0x20,
//P
CR00
0xbb,
CR01
0x8F,
CR02
0x8F,
CR03
0x9f,
CR04
0x98,
CR05
0x87,
CR06
0x83,
CR07
0x10,
CR09
0x40,
CR10
0x60,
CR11
0x03,
CR12
0x5F,
CR13
0x90,
CR15
0x5F,
CR16
0x84,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x00,
0x22107000,
0x22210000,
0x22415000
466
//99-MHz dot clock
®
R
0x22107000,
0x22210000,
0x22415000
MSR mask
0x00
467
®
R
1152x864_75Hz =
Dot clock value
108,
M value
0x0010,
//M
N value
0x0002,
//N
P value
0x20,
//P
CR00
0xC3,
CR01
0x8F,
CR02
0x8F,
CR03
0x87,
CR04
0x97,
CR05
0x07,
CR06
0x82,
CR07
0x10,
CR09
0x40,
CR10
0x60,
CR11
0x03,
CR12
0x5F,
CR13
0x90,
CR15
0x5F,
CR16
0x83,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x22107000,
0x22210000,
0x22415000,
468
//108-MHz dot clock
®
R
0x22107000,
0x22210000,
0x22415000,
MSR mask
0x00
469
®
R
1152x864_85Hz =
Dot clock value
121,
M value
0x006D,
//M
N value
0x0014,
//N
P value
0x20,
//P
CR00
0xc0,
CR01
0x8F,
CR02
0x8F,
CR03
0x84,
CR04
0x97,
CR05
0x07,
CR06
0x93,
CR07
0x10,
CR09
0x40,
CR10
0x60,
CR11
0x03,
CR12
0x5F,
CR13
0x90,
CR15
0x5F,
CR16
0x94,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x2220C000,
0x22210000,
0x22415000,
470
//121-MHz dot clock
®
R
0x2220C000,
0x22210000,
0x22415000,
MSR mask
0x0
471
®
R
1280x720_60Hz =
Dot clock value
74,
M value
0x0023,
//M
N value
0x0004,
//N
P value
0x30,
//P
CR00
0xCB,
CR01
0x9F,
CR02
0xA0,
CR03
0x8E,
CR04
0xB3,
CR05
0x04,
CR06
0xE8,
CR07
0xF0,
CR09
0x60,
CR10
0xD0,
CR11
0x03,
CR12
0xCF,
CR13
0xA0,
CR15
0xD0,
CR16
0xE8,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x2210A000,
0x22210000,
0x22415000,
472
//74-MHz dot clock
®
R
0x2210A000,
0x22210000,
0x22415000,
MSR mask
0x00
473
®
R
1280x720_75Hz =
Dot clock value
96,
M value
0x000A,
//M
N value
0x0001,
//N
P value
0x20,
//P
CR00
0xCF,
CR01
0x9F,
CR02
0xA0,
CR03
0x92,
CR04
0xB3,
CR05
0x04,
CR06
0xEE,
CR07
0xF0,
CR09
0x60,
CR10
0xD0,
CR11
0x03,
CR12
0xCF,
CR13
0xA0,
CR15
0xD0,
CR16
0xEE,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x2210A000,
0x22220000,
0x22419000,
474
//96-MHz dot clock
®
R
0x2210A000,
0x22220000,
0x22419000,
MSR mask
0x00
475
®
R
1280x720_85Hz =
Dot clock value
110,
M value
0x0035,
//M
N value
0x000A,
//N
P value
0x20,
//P
CR00
0xD1,
CR01
0x9F,
CR02
0xA0,
CR03
0x94,
CR04
0xB3,
CR05
0x04,
CR06
0xF2,
CR07
0xF0,
CR09
0x60,
CR10
0xD0,
CR11
0x03,
CR12
0xCF,
CR13
0xA0,
CR15
0xD0,
CR16
0xF2,
CR30
0x02,
CR31
0x02,
CR32
0x02,
CR33
0x02,
CR35
0x00,
CR39
0x00,
0x2210A000,
0x22220000,
0x2241B000,
476
//110-MHz dot clock
®
R
0x2210A000,
0x22220000,
0x2241B000,
MSR mask
0x00
477
®
R
1280x960_60Hz =
Dot clock value
108,
M value
0x0010,
//M
N value
0x0002,
//N
P value
0x20,
//P
CR00
0xDC,
CR01
0x9F,
CR02
0x9F,
CR03
0x80,
CR04
0xAB,
CR05
0x99,
CR06
0xE6,
CR07
0x10,
CR09
0x40,
CR10
0xC0,
CR11
0x03,
CR12
0xBF,
CR13
0xA0,
CR15
0xBF,
CR16
0xE7,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x2210A000,
0x22210000,
0x22415000,
478
//108-MHz dot clock
®
R
0x2210A000,
0x22210000,
0x22415000,
MSR mask
0x00
479
®
R
1280x960_75Hz =
Dot clock value
129,
M value
0x0029,
//M
N value
0x0006,
//N
P value
0x20,
//P
CR00
0xD3,
CR01
0x9F,
CR02
0x9F,
CR03
0x97,
CR04
0xaa,
CR05
0x1b,
CR06
0xE8,
CR07
0x10,
CR09
0x40,
CR10
0xC0,
CR11
0x03,
CR12
0xBF,
CR13
0xA0,
CR15
0xBF,
CR16
0xE9,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x2210A000,
0x22210000,
0x2241B000,
480
//129-MHz dot clock
®
R
0x2210A000,
0x22210000,
0x2241B000,
MSR mask
0x00
481
®
R
1280x960_85Hz =
Dot clock value
148,
M value
0x0042,
//M
N value
0x0009,
//N
P value
0x20,
//P
CR00
0xD3,
CR01
0x9F,
CR02
0x9F,
CR03
0x97,
CR04
0xA7,
CR05
0x1B,
CR06
0xF1,
CR07
0x10,
CR09
0x40,
CR10
0xC0,
CR11
0x03,
CR12
0xBF,
CR13
0xA0,
CR15
0xBF,
CR16
0xF2,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x2210A000,
0x22220000,
0x2241D000,
482
®
R
0x2210A000,
0x22220000,
0x2241D000,
MSR mask
0x00
483
®
R
1280x1024_60Hz =
Dot clock value
108,
M value
0x0010,
//M
N value
0x0002,
//N
P value
0x20,
//P
CR00
0xCE,
CR01
0x9F,
CR02
0x9F,
CR03
0x92,
CR04
0xA5,
CR05
0x13,
CR06
0x28,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0xA0,
CR15
0xFF,
CR16
0x29,
CR30
0x04,
CR31
0x03,
CR32
0x04,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x22107000,
0x22210000,
0x22415000,
484
//108-MHz dot clock
®
R
0x22107000,
0x22210000,
0x22415000,
MSR mask
0x00
485
®
R
1280x1024_70Hz =
Dot clock value
129,
M value
0x0029,
//M
N value
0x0006,
//N
P value
0x20,
//P
CR00
0xD3,
CR01
0x9F,
CR02
0x9F,
CR03
0x97,
CR04
0xAA,
CR05
0x1B,
CR06
0x28,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0xA0,
CR15
0xFF,
CR16
0x29,
CR30
0x04,
CR31
0x03,
CR32
0x04,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x22107000,
0x22210000,
0x22419000,
486
®
R
0x22107000,
0x22210000,
0x22419000,
MSR mask
0x00
487
®
R
1280x1024_72Hz =
Dot clock value
132,
//132-MHz dot clock
//notinvesa
M value
0x0014,
//M
N value
0x0002,
//N
P value
0x20,
//P
CR00
0xd3,
CR01
0x9F,
CR02
0x9F,
CR03
0x97,
CR04
0xAa,
CR05
0x1b,
CR06
0x29,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0xA0,
CR15
0xFF,
CR16
0x2a,
CR30
0x04,
CR31
0x03,
CR32
0x04,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x22109000,
0x22314000,
0x22515000, //224517000
488
®
R
0x22109000,
0x22314000,
0x22515000, //224517000
MSR mask
0x00
489
®
R
1280x1024_75Hz =
Dot clock value
135,
M value
0x002B,
//M
N value
0x0006,
//N
P value
0x20,
//P
CR00
0xCE,
CR01
0x9F,
CR02
0x9F,
CR03
0x92,
CR04
0xA1,
CR05
0x13,
CR06
0x28,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0xA0,
CR15
0xFF,
CR16
0x29,
CR30
0x04,
CR31
0x03,
CR32
0x04,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x22109000,
0x22314000,
0x2251D000,
490
//135-MHz dot clock
®
R
0x22109000,
0x22314000,
0x2251C000,
MSR mask
0x00
491
®
R
1280x1024_85Hz =
Dot clock value
157,
M value
0x0050,
//M
N value
0x0017,
//N
P value
0x10,
//P
CR00
0xD3,
CR01
0x9F,
CR02
0x9F,
CR03
0x97,
CR04
0xA7,
CR05
0x1B,
CR06
0x2E,
CR07
0x10,
CR09
0x40,
CR10
0x00,
CR11
0x03,
CR12
0xFF,
CR13
0xA0,
CR15
0xFF,
CR16
0x2F,
CR30
0x04,
CR31
0x03,
CR32
0x04,
CR33
0x03,
CR35
0x00,
CR39
0x01,
0x2210B000,
0x22415000,
0x2251D000,
492
®
R
0x2210B000,
0x22415000,
0x2251D000,
MSR mask
0x00
493
®
R
1600x900_60Hz =
Dot clock value
119,
M value
0x006B,
//M
N value
0x0014,
//N
P value
0x20,
//P
CR00
0x05,
CR01
0xC7,
CR02
0xC8,
CR03
0x88,
CR04
0xDF,
CR05
0x14,
CR06
0xA2,
CR07
0xFF,
CR09
0x60,
CR10
0x84,
CR11
0x07,
CR12
0x83,
CR13
0xC8,
CR15
0x84,
CR16
0xA2,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x01,
CR39
0x00,
0x2220E000,
0x22416000,
0x44419000,
494
//Dot clock
®
R
0x2220E000,
0x22416000,
0x44419000,
MSR mask
0x00
495
®
R
1600x900_75Hz =
Dot clock value
152,
M value
0x0011,
//M
N value
0x0004,
//N
P value
0x10,
//P
CR00
0x09,
CR01
0xC7,
CR02
0xC8,
CR03
0x8C,
CR04
0xE0,
CR05
0x16,
CR06
0xAA,
CR07
0xFF,
CR09
0x60,
CR10
0x84,
CR11
0x07,
CR12
0x83,
CR13
0xC8,
CR15
0x84,
CR16
0xAA,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000,
0x44419000,
496
//Dot clock
®
R
0x2220e000,
0x22416000,
0x44419000,
MSR mask
0x00
497
®
R
1600x900_85Hz =
Dot clock value
175,
M value
0x0031,
//M
N value
0x000C,
//N
P value
0x10,
//P
CR00
0x0B,
CR01
0xC7,
CR02
0xC8,
CR03
0x8E,
CR04
0xE0,
CR05
0x16,
CR06
0xAF,
CR07
0xFF,
CR09
0x60,
CR10
0x84,
CR11
0x07,
CR12
0x83,
CR13
0xC8,
CR15
0x84,
CR16
0xAF,
CR30
0x03,
CR31
0x03,
CR32
0x03,
CR33
0x03,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000,
0x44419000,
498
//Dot clock
®
R
0x2220e000,
0x22416000,
0x44419000,
MSR mask
0x00
499
®
R
1600x1200_60Hz =
Dot clock value
162,
M value
0x0019,
//M
N value
0x0006,
//N
P value
0x10,
//P
CR00
0x09,
CR01
0xC7,
CR02
0xC7,
CR03
0x8D,
CR04
0xcf,
CR05
0x07,
CR06
0xE0,
CR07
0x10,
CR09
0x40,
CR10
0xB0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xAF,
CR16
0xE1,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x2210b000,
0x22416000,
0x44419000,
500
//162-MHz dot clock,
®
R
0x2210b000,
0x22416000,
0x44419000,
MSR mask
0x00
501
®
R
1600x1200_65Hz =
Dot clock value
175,
M value
0x005d,
//M
N value
0x0018,
//N
P value
0x10,
//P
CR00
0x09,
CR01
0xC7,
CR02
0xC7,
CR03
0x8D,
CR04
0xcf,
CR05
0x07,
CR06
0xE0,
CR07
0x10,
CR09
0x40,
CR10
0xB0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xAF,
CR16
0xE1,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x2210c000,
0x22416000,
0x44419000,
502
//175.5-MHz dot clock,
®
R
0x2210c000,
0x22416000,
0x44419000,
MSR mask
0x00
503
®
R
1600x1200_70Hz =
Dot clock value
189,
M value
0x003D,
//M
N value
0x000e,
//N
P value
0x10,
CR00
0x09,
CR01
0xC7,
CR02
0xC7,
CR03
0x8d,
CR04
0xcf,
CR05
0x07,
CR06
0xE0,
CR07
0x10,
CR09
0x40,
CR10
0xb0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xaf,
CR16
0xE1,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000
0x44419000
504
//189-MHz dot clock
//P
®
R
0x2220e000,
0x22416000
0x44419000
MSR mask
0x00
505
®
R
1600x1200_72Hz =
Dot clock value
195,
M value
0x003f,
//M
N value
0x000e,
//N
P value
0x10,
CR00
0x0b,
CR01
0xC7,
CR02
0xC7,
CR03
0x8f,
CR04
0xd5,
CR05
0x0b,
CR06
0xE1,
CR07
0x10,
CR09
0x40,
CR10
0xb0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xaf,
CR16
0xe2,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000,
0x44419000,
506
//195-MHz dot clock
//P
®
R
0x2220e000,
0x22416000,
0x44419000,
MSR mask
0x00
507
®
R
1600x1200_75Hz =
Dot clock value
202,
M value
0x0024,
//M
N value
0x0007,
//N
P value
0x10,
//P
CR00
0x09,
CR01
0xC7,
CR02
0xC7,
CR03
0x8d,
CR04
0xcf,
CR05
0x07,
CR06
0xE0,
CR07
0x10,
CR09
0x40,
CR10
0xb0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xaf,
CR16
0xE1,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000,
0x44419000,
508
®
R
0x2220e000,
0x22416000,
0x44419000,
MSR mask
0x00
509
®
R
1600x1200_85Hz =
Dot clock value
229,
M value
0x0029,
//M
N value
0x0007,
//N
P value
0x10,
//P
CR00
0x09,
CR01
0xC7,
CR02
0xC7,
CR03
0x8d,
CR04
0xcf,
CR05
0x07,
CR06
0xE0,
CR07
0x10,
CR09
0x40,
CR10
0xb0,
CR11
0x03,
CR12
0xAF,
CR13
0xC8,
CR15
0xaf,
CR16
0xE1,
CR30
0x04,
CR31
0x04,
CR32
0x04,
CR33
0x04,
CR35
0x01,
CR39
0x00,
0x22210000,
0x22416000,
0x0,
510
®
R
0x22210000,
0x22416000,
0x0,
MSR mask
0x00
511
®
R
1792x1344_60Hz =
Dot clock value
204,
M value
0x003e,
//M
N value
0x000d,
//N
P value
0x10,
//P
CR00
0x2d,
CR01
0xdf,
CR02
0xdf,
CR03
0x91,
CR04
0xEf,
CR05
0x88,
CR06
0x70,
CR07
0x10,
CR09
0x40,
CR10
0x40,
CR11
0x03,
CR12
0x3F,
CR13
0xC8,
CR15
0x3f,
CR16
0x71,
CR30
0x05,
CR31
0x05,
CR32
0x05,
CR33
0x05,
CR35
0x01,
CR39
0x00,
0x2220e000,
0x22416000,
0x44419000,
512
®
R
0x2220e000,
0x22416000,
0x44419000,
MSR mask
0x40
513
®
R
1856x1392_60Hz =
Dot clock value
218,
M value
0x0030,
//M
N value
0x0009,
//N
P value
0x10,
//P
CR00
0x37,
CR01
0xe7,
CR02
0xe7,
CR03
0x9b,
CR04
0xf3,
CR05
0x8f,
CR06
0x9d,
CR07
0x10,
CR09
0x40,
CR10
0x70,
CR11
0x03,
CR12
0x6f,
CR13
0xC8,
CR15
0x6f,
CR16
0x9e,
CR30
0x05,
CR31
0x05,
CR32
0x05,
CR33
0x05,
CR35
0x01,
CR39
0x00,
0x2220f000,
0x0,
0x0,
514
®
R
0x2220f000,
0x0,
0x0,
MSR mask
0x40
515
®
R
1920x1440_60Hz =
Dot clock value
234,
M value
0x0025,
// M
N value
0x0006,
// N
P value
0x10,
// P
CR00
0x40,
CR01
0xef,
CR02
0xef,
CR03
0x84,
CR04
0xff,
CR05
0x19,
CR06
0xda,
CR07
0x10,
CR09
0x40,
CR10
0xa0,
CR11
0x03,
CR12
0x9f,
CR13
0xf0,
CR15
0x9f,
CR16
0xdb,
CR30
0x05,
CR31
0x05,
CR32
0x05,
CR33
0x05,
CR35
0x01,
CR39
0x00,
0x22210000,
0x0,
0x0,
516
// 234-MHz dot clock
®
R
0x22210000,
0x0,
0x0,
MSR mask
0x40
517
®
R
1856x1392_60Hz =
Dot clock value
218,
M value
0x0030,
// M
N value
0x0009,
// N
P value
0x10,
// P
CR00
0x37,
CR01
0xe7,
CR02
0xe7,
CR03
0x9b,
CR04
0xf3,
CR05
0x8f,
CR06
0x9d,
CR07
0x10,
CR09
0x40,
CR10
0x70,
CR11
0x03,
CR12
0x6f,
CR13
0xC8,
CR15
0x6f,
CR16
0x9e,
CR30
0x05,
CR31
0x05,
CR32
0x05,
CR33
0x05,
CR35
0x01,
CR39
0x00,
0x2220f000,
0x0,
0x0,
518
// 218.25-MHz dot clock
®
R
0x2220f000,
0x0,
0x0,
MSR mask
0x40
519

Mirror FTP Sites

Transcription

Similar documents

Evaluating the security of consumer devices

bits and pieces

Brochure

TouchSmart tm2-2050us Notebook PC

Intel® Communications Chipset 8900 to 8920 Series Software

stream your

- CUNY Academic Works - The City University of New York

Mar. 2013 - Filipino Catholic Chaplaincy in Vienna

pc gaming!

Product Sheet - Projector Central

reviews - Hardware

a careful eye for the caring professions smart health-care

Intel® Fortran Compiler for Linux* Systems User`s Guide, Volume II

AMD64 Architecture Programmer`s Manual, Volume 2